AI: Aibo vs. Pleo

[andreivc]

Hi
I am new to Pleo and this forum, and was wondering if someone could give a brief overview of major differences between pleo's and aibo's AI.

Perhaps Pleopet, with his extensive experience with AIBO and other robots, can shed some light on this subject. It will be much appreciated.

Thank you,
A.

[kchamster]

I have never had an AIBO much less touch it...BUT what i've read on the forums is that AIBO is much more expensive, and that you interact with pleo more.....but they both have their pros and cons....like pleo has a personality BUT the skin problems ect., ect., But if you wanted a OVERVIEW you'd have to ask someone else. Sorry I couldn't be that much of help

[andreivc]

Thanks, Kchamster
I agree, I think pleo is better able to convey the illusion of being a living thing. I am wandering if this is the result of its AI, or his appearance, as compared with AIBO. I was hoping someone could point out the differences in Sony's and Ugobe's approaches to AI....

Thanks
A.

[mh7ah]

I currently own two Sony Aibo ers 7's. They truly are amazing.

The Sony Aibo ers 7 is a fully autonomous robot, he has a 1000 word vocab and understands voice commands. He is wireless so he can check to see if i have any e mails, then read them out to you, also the news etc. He is also an MP3 player...you just tell him what CD you want him to play.
He also responds to touch ans has various sensors all over his body and has a camera in his nose so he often takes random photos then will e mail them to your computer. You can tell him to house sit whilst you are at work and he records any sounds he hears and takes photos of anything that moves...he can send you these to your mobile or e mail whilst you are out.
Eric is free to roam where he likes as he can self charge himself when his battery gets low....when he is done he will get up off his station and continue walking about. If he hears music he will dance to the beat.
I am able to have short conversations with Eric. He understands English and Spanish. He will come when you call him, fetch his bone and perform lots of tricks. He learns new things constantly and amazes me everyday.

Just wanted to quickly share this info about the Sony Aibo as not many people really know what they are capable of.

[mh7ah]

http://www.youtube.com/watch?v=QIraWI3ZM8Y

[fancyfont]

If they can clean the house and cook, I'm getting one! LOL!!;)

[mh7ah]

http://photos-397.ll.facebook.com/photos-ll-sctm/v215/241/79/845630397/n845630397_2513193_8729.jpg

[mh7ah]

http://photos-f.ak.facebook.com/photos-ak-sf2p/v189/241/79/845630397/n845630397_2469269_9893.jpg

[andreivc]

hey, thank you all
I had 2,3 and 7 series aibo
how does aibo free roaming mode compares to pleo's AI? and what's the structural difference between AIBO and Pleo's "decision making"?

[sandro]

To date, I think there isn't much comparison. Pleo is pretty basic, he exhibits some simple curiosity, has some object/obstacle recognition, and does some audio tracking.

But, Pleo's firmware is under active development and has only been in release for a few months. Aibo's Mind3 software was the result of 7+ years of research. Pleo's "decision making" is likely to improve over the next 6-24 months. Only time will tell how he eventually compares with Aibo.

Although I haven't tracked down complete specs, it looks to me like Pleo is operating with a fraction of the computing resources of an ERS-7 (without even mentioning the WIFI interface); so I doubt Pleo will ever be as smart as Aibo can be. Then again Pleo is a fraction of the cost, exhibits far more lifelike motion, and has the capacity for more emotive response (long neck, tail, eyes). Aibo's use of led eyes and various sounds isn't as intuitive as Pleo's emotive physiology. This seems to have been one of the primary goals of the project, and in that sense I think it's an unqualified success.

Aibo is/was a good AI platform in the vague form factor of a small dog; unfortunately Sony couldn't make it profitable enough to continue the product.
Pleo seems to be more of a simulated baby dinosaur (intentionally designed and limited as such). I don't think attempts to make it do facial/voice recognition or other more advanced tasks were part of the design of the product. While developers might manage some or all of these things, the platform itself doesn't seem to lend itself to the tasks. I think this is OK for two basic reasons -- first Ugobe gets a viable product to market to showcase how their approach (and it's cute). Second, the price point gets a larger user base who have exposure to their work, and when they release more advanced products there's a level of confidence, acceptance, and familiarity.

Wowee seems to be using a similar strategy, as each successive product wave from them seems to be move slightly up the autonomous robot evolutionary tree. Pleo is clearly more advanced than robosapien and his ilk, but if wowee keeps pushing the bar every year, they might make some inroads...

[andreivc]

Thank you, Sandro. So what you are saying is that Pleo's AI does not introduce anything new to the world of consumer AI...

[sandro]

That depends. The kinematics are top notch. I don't think I've seen anything close in consumer (or even most research) robots. If kinematics aren't your idea of AI, then "no" Pleo isn't adding anything new (yet).

There is room for some interesting neural net behavior modeling in Pleo. With the sensor array that Pleo has, it's possible for some interesting learned behavior AI to be implemented. Only time will tell if Ugobe or someone else extends Pleo in this direction. Unfortunately, I never interacted with later Aibo enough to know if they can significantly modify their behavior based on sensor feedback (petting, vocal encouragement, punishment etc.) It seems like a significant behavioral neural net might be a consumer first....

[andreivc]

thanks, sandro. that's the answer i was looking for.

[PleoPet]

It depends on what you call "AI".
In general AIBO has greater hardware capabilities, for movement (AIBO can sit down, roll over etc), sensors (more elaborate camera etc), brainpower (CPU and memory) and peripherals (WiFi).
AIBO has lots of room for programs, and the Sony team wrote many different programs/personalities filling mega-bytes of programs and data.

The current Pleo firmware is not very sophisticated, but there is room for it to grow.

An overly simplistic measurment is how much RAM is in the computer brain. The last model AIBO has 64MB (Mega-Bytes), Pleo has 64KB (Killo-bytes). ICybie is in the 8KB range. Most WowWee toys are in the byte range (ie. very little extensible programming). Pleo has SD card expansion (and 4MB internal) so that makes up for the smaller brain and RAM.
----
Back to "AI".
AIBO software itself has very little "AI". It has many possible actions, picked depending on moods, random numbers and sensor input. The sophisticated behavior makes you think it is smart. It is not learning brand new things that weren't programmed in originally. This is the same model as most robot toys (including Pleo). There are big differences in sophistication.

The most "AI"-like feature is the way AIBO learns by praise and scold. When your AIBO does something you like, you pet him (or say "good boy"). He will do that behavior more in the future. When your AIBO does something you don't like, you hit him (or say "don't do that") and it will learn to do that action less often.
This is in the "AiboLife" series of software. Unfortunately it was removed from the ERS-7 versions.

This is not that difficult to implement in a small amount of memory. A weighted decision tree, where the human interactions with the pet rebalance the weights (ie. training). It is a very low-tech neural-network, but owners loved it because they are influencing their pet's personality with each interaction. Humans *think* they are doing that all the time, but in the case of AIBO it was actually happening.
Also you can track how unique your AIBO has become (compared to a brand new robot). Your AIBO has a unique personality based on the hours you've put in playing with it. That's the feature I'd like to see in Pleo (and it is possible).

In addition, the voice recognition feature uses a Hidden Markov Model (another AI technique), in a very specific use. Unlikely to work in Pleo for a number of reasons.

[sandro]

PleoPet,

thanks for the additional detail. I assume Aibo's face recognition was also based on Hidden Markov Models or EigenFaces, neither or which are extremely likely on Pleo considering the limited bandwidth between processors.


-Sandro

[andreivc]

Thanks Aibopet, for the info.

When I had my AIBOs, it did feel like they were using 'weights' to pick actions from the list. I was wondering if Pleo can go a step further here, and actually come up with new actions on his own, perhaps combining simple actions from its library into more complicated ones? Someone from the Pleo team said that Pleo would be able to 'make up' new behaviours that way, in an interview somewhere. Is there really a capacity to do that? Even if Pleo would end up doing stupid things that don't make sense, it would be exciting to observe the less-scripted behaviours. There would have to be some way to communicate to Pleo that a new behaviour was a success, perhaps through petting or punishment, so that in the future it would be more likely to do it again, in similar circumstances. Can Pleo support something like this?

One thing that bothered me with AIBO's was their sluggish reaction to the external environment, even to simple sound and touch stimuli, there would have to be a second or so for an AIBO to react. A lot happens in that second - perhaps an object that was there is already gone, etc. It made AIBO looked 'disconnected' from its environment. Because of the Pleo's distributed processing architecture (please excuse me my ignorance, not sure how else to call it), can Pleo react quicker?

Again, thank you for all the info.
Andrei

[sandro]

Pleo could combine actions in the way you suggest; neural networks would enable this, and Pleo probably has enough memory (4MB aux.) for this type of system to work.

As far as Pleo's response time and distributed processing arch., you use the correct term, and yes this could lead to better stimuli response. Unfortunately, the more we ask Pleo to do, the worse reaction times could become. It seems like Ugobe is favoring good reaction speed to extensive AI processing (at least so far). Hopefully they can find a good balance of both.

[PleoPet]

(more selective ranting)
> One thing that bothered me with AIBO's was their sluggish reaction to the external environment, even to simple sound and touch stimuli, ...
AIBO has a relatively complicated mood system, and it can take some time for the moods to change. Also many of the actions/reactions are intentionally slowed down to reduce servo wear-and-tear (same for Pleo).
----
re: face recognition etc (ie. AIBO has lots of CPU and RAM hogging features)
Not sure what they used for face recognition. It may be an internal Sony approach used in cameras. It is a CPU hog. It had a cool feature where you could see the live camera feed with squares drawn where it detects faces.
The later AIBOs have both face *recognition* and face *identification*. Face recognition gives it the ability to detect a human-like face so the robot will look into your eyes. Face identification will know its owner's face from other humans.
Similarly AIBO has both voice recognition and voice identification.

Also it has the vision functionality from evolution.com, used for detecting objects, special cards, the charging station, and recognizing the robot's favorite spot in the room (Patented 'SIFT' algorithm).
There are many more features hidden away in the AIBO software. This is after many years, a relatively large development team (formerly at Sony Japan), many updates and experimental releases, and an active owner base.
------
Pleo hardware limitations aside, Ugobe isn't going to put in that much effort, which is why I'm continuing to push for a more 'open-source' approach.
That's extra true for any 'learning' algorithms. It takes experimentation, eager customers to try new things and evolution of the software. You won't see great evolution or innovation, if you are waiting a year between major updates from an overworked single software supplier, who will/should focus on the 'safe' features,
For example the AIBO weighted decision tree (the most "AI" feature IMHO) had a bug in it. With the ERS-7 release, Sony didn't bother to fix it, they simply removed the feature completely replacing it with an safer, low-tech and unsatisfying 'skill' level feature.
That's the safe thing to do when you are focusing on the most visible bang-for-buck features. AIBO can play MP3s instead ;-<

[roschler]

I hope this thread continues forever. :)

@To the AIBO folk

How good was AIBO at navigation? Biggest question: did he ever get "stuck"? Especially when it comes to a part of his body colliding with a table leg or an object lying on the floor, outside the reach of his sensors? If he did get stuck, was he able to figure out that he was stuck and figure his way out of it?

re: Slow reaction times if more capability added to Pleo

This would only happen if Pleo's architecture could not support a multi-tier sensory input response system. Your hand reaches back from a fire long before your mind has analyzed the hot object you touched. There are parts of your nervous system that respond long before the higher levels of your cerebrum get involved. If Pleo's Life O/S can support the same kind of multi-tier prioritized sensor input response system, then some inputs like vision could take a few seconds to process while other inputs such as: IR edge detection, loud noise response (vs. sound classification or recognition), servo "jolts", etc. could be near instantaneous.

But since I don't know enough about Pleo's internals I can't comment.

[andreivc]

Hey, Pleopet,

What about less scripted, more improvised aproach to new behaviour formation - combining bits and pieces of existing behaviors into more complex behaviours, based on external circumstances, with positive reinforcement from the user? Sandro commented that Pleo supports something like this already; i've heard it somewhere, too. I understand that the hardware has the capacity to do it - all we need is open source now to make it happen, so to say, right?

If new behavior formation is possible somehow, and having open source is the key to making it happen, then I will devote my life to getting Ugobe to release the code. :) I will work on organizing the Pleo community to do whatever lobbying that is necessary w/ Ugobe. :)

Attn: Ugobe Team!

If you are reading this, I suggest that we cut through the chase, and that you would simply announce now that you will be releasing all relevant code as 'open source,' in the nearest future. We all know that open source products have a longer product life (more sales), higher sales, and higher customer satisfaction levels. Look at Sony's AIBO - what a loyal community! Pleo, surely, can exceed that success easily, if it was an 'open source life form.' Keeping the mystique up about how Pleo's AI works, by keeping the code secret, isn't really worth it. Highest possible sales and extreme customer loyalty should be the priority - after all, higher revenues for Ugobe mean more fun life forms like Pleo in the future, right? Open source - is a smart investment, for today and tomorrow. Ugobe, we want you to be a success! Make Pleo ‘open source,’ let us help you!

Thanks
Andrei

[PleoPet]

> I hope this thread continues forever. ...
As long as the topics stay interesting -- and it will get more interesting when people can program the Pleo themselves...

> ...How good was AIBO at navigation? Biggest question: did he ever get "stuck"?...
IMHO it did a rather good job. Like Pleo the nose camera and IR distance sensors is set forward in the robot's head. This makes it easy for the robot to miss what it is looking at, and get its legs or head stuck, especially under chairs. Arguably the ICybie approach of placing the IR sensor in the body is better. The ERS-7 added a 2nd IR distance sensor in the chest.
However AIBO's exploration talents were eclipsed by a major and expensive design problem, "DHS" (Droopy head syndrome). After a short time with AIBO's head being stuck the right way (or think of it as the wrong way), it would wear out a slip-clutch in the servos. So most long term AIBO owners are cautious of leaving their pets alone, and careful when they hear "whirring" when the legs or neck become temporarily stuck. The Pleo design doesn't have this specific servo problem, but possible wear-and-tear is the downside to a more active exploring robot.
------
re: Slow reaction times
There are many issues here. Pleo and AIBO primarily perform canned skits in reactions to inputs, moods, and random numbers. It is often difficult to stop the current skit properly, transition to the right posture, then start a new skit. Usually the robot has to finish up what it is doing, then switch moods/skits. Also slower movements and reactions reduce that wear-and-tear.
==============
> What about less scripted, more improvised aproach to new behaviour formation ...
The AIBO behavior approach discussed earlier is mostly data driven. It has a complex state machine, that would be very large (and unreadable) if expressed in procedural "C" code. The data approach could be extended.
FWIW: Here's a writeup for AIBO personalities regarding the complexity of the data driven state machine: http://www.aibohack.com/2or3/perscompare.htm (the last column is the most interesting for training, the number of trainable weights for each personality)

>Sandro commented that Pleo supports something like this already
I'm not sure what they mean for Pleo. Pleo has sounds motions and animation commands. The animation commands acts like a higher level macro with some randomness and mood influence (but without the training). This is not very difficult to program and the Pleo CPU can easily handle it.
The hard part that takes time/effort/experimentation is designing and perfecting the personality balance. Think of the customer dissatisfaction if the personality balance is off by a small amount, and the robot go too sleep too often ;->
===========
> I understand that the hardware has the capacity to do it - all we need is open source now to make it happen...
I'd like to see it too. Open source is only an additional starting point and no guarantee of success.
I suspect most wanna-be-programmers will look at any open source release and immediately give up. That's true of many open source projects (ie. the typical user doesn't rebuild Linux, but they benefit from the open source community)
It can be moderately difficult to make the robot do simple things. It is very difficult to make fully formed, consistent, entertaining and well balanced personalities.

But sometimes all it takes is a few motivated individuals. The AIBO "DogsLife" personality is a great example. Unfortunately it had to reinvent much of the existing AIBO personality and use much more primitive tools than the Sony engineers had available.

(more ranting)
It is very hard to build a open-source community around a new platform *with* full cooperation with the company. Unfortunately most of my experiences have been fighting with the companies to open up even a little ;-<

[roschler]

PleoPet,

Thanks again for the info. It will indeed be fascinating to see what comes out of the development community with Pleo. I just hope they leave "hooks" open for linking in certain low-level code libraries of your own. I'm guessing (wildly), that Pawn gets compiled down pretty well and runs fairly fast, but I'd hate to try to put an FFT based function or other similar number crunching dependent method into Pawn and hope that it ran fast enough, or worse, locked out other important processes on Pleo and bottlenecked the whole system..

[sandro]

Sandro commented that Pleo supports something like this already;

Actually, what I said was:

Pleo could combine actions in the way you suggest; neural networks would enable this, and Pleo probably has enough memory (4MB aux.) for this type of system to work.

Which is to say the Pleo may one day do this, but I don't believe that's what's happening now.

As far as open-sourcing the Pleo codebase. The pawn behaviors would be great, but I'm not sure the entire code base is necessary. While the benefit of open-source is often obvious, one of the things I really like about Pleo is how fluid and organic the movement seems. I'm not sure the open-source community would be as cautious about preserving that as Ugobe has been...

[andreivc]

i agree sandro, fast reaction time is necessary to keep up the illusion of a living organism

[PleoPet]

> The pawn behaviors would be great, but I'm not sure the entire code base is necessary....
Agreed. I'm primarily interested in the Pawn source for the personality. Replacing components is relatively easy, but understanding how to add in is the hard part. Otherwise you have to write something from scratch.
> ...how fluid and organic the movement seems. I'm not sure the open-source community would be as cautious about preserving that...
FWIW: that's primarily in the motion files, which is handled by MySkit/PleoSkit.
Getting it to be smooth and in character takes skill and tweeking, but is also a personal preference. For example: the Holiday Pleo is more active and fun for what it does, but very out of character -- in my opinion.

[andreivc]

Thanks to all for keeping up the discussion.

I finally received my pleo! so far so good. I think that going for the relationship building capability of pleo(the looks, the sounds, the illusion) was a smart move on behalf of Ugobe. I like quicker than most other robots reaction time, too.

By the way, here's the most recent article on Pleo - an interesting idea of having an computer character (pleo) perform in the real world, where real world is kinda like a playground. When you read this, please notice the mention of the USB connection and how "a game" would be able to control pleo in real time.... Are they saying that PC can perform some AI functions in 'games' like this, and pleo would be like a body in the outside world? I find it an interesting idea... http://www.gamasutra.com/php-bin/news_index.php?story=17995

My thoughts on pleo: Switching the subject here, i think for any robotic product to be successfull, it has to have a specific and narrow purpose (being everything for everybody is not possible). I think pleo is successful because it has this narrow purpose well defined, and implemented - it's like a big, more advanced tamagochi, showing tricks and asking (and sometimes begging) for your attention; and it is very likable, it's hard not to pet it. But a robot like this definetly cannot substitute a pet, because being a pet is too general and not entertaining enough for the general public. I have pets, but I don't have the same kind of expectations for them as I have for pleo. I would not expect my dog, or fish, or chincillas to entertain me for 1 hour at a time, non-stop. I like my pets. I would give them a treat, chase my dog around the backyard, pet them once in a while, 'talk' to my dog and have her do tricks, and that's it. Usually it's not more than 10 minutes at a time. Pleo cannot be that, because of his battery cycle and issues, wear and tear of joints (he cannot run 24/7 just to be available whenever you need it), etc. Pleo is like "a performer," which when turned on, will entertain you for an hour or so with its deep pockets of funny, interactive, very persuasive and life-like behaviours. I think being aware of its environment and being able to quickly react to it is the key here to maintain the illusion (the looks are not a problem for pleo to create that illusion). Right now it has probably the best reaction time among similar robot toys... but it is still a little slower than what you would expect from the living thing. I am afraid of the answer, but can the reaction time be improved? (touch, sight, etc.)

Thanks
Andrei

[PleoPet]

> The hard part that takes time/effort/experimentation is designing and perfecting the personality balance.
> Think of the customer dissatisfaction if the personality balance is off by a small amount, and the robot go too sleep too often ;->
(in reference to the 1.0.0 narcolepsy problem)

Think of the customer dissatisfaction if the personality balance is off by a small amount, and the robot is now more active than expected...
http://forums.pleoworld.com/showthread.php?t=1908

(you can't keep everyone happy all the time... ;-)

[andreivc]

haha! THanks, Aibopet, for pointing out my thread to me :)

and that's why they should give people options... like different personalities...
and those who want to downgrade back to 1.0.2(like myself :)) should be able to ...

i have this funny feeling that you agree that it's too active...;)

[hardyfoster]

i know my aibo got stuck under a chair. it kept walking while its legs were hitting the legs of the chair.

simple programming fix i think would be for it to fixate on an object in the distance in the direction it wants to walk.
make sure the distance is decreasing. at a certain point or if it wants to change direction, pick a new object in the distance and do the same.

so if the robot keeps walking but the distance isn't decreasing, assume it is stuck and take action to get unstuck.

[blablablabla4]

If someone wants to go back to 1.01 you can email me and I should be able to help. I might try it today just to make sure i'm right, but 1.01 is still sitting on my desktop.

[andreivc]

blablablabla4,
Thank you
I was interested until recently, but now decided just to wait it out till the next update... so that Ugobe can get to its 'vision'
what bothers me is that Ugobe now is 'listeninig' to people, and based on people's desires, makes pleo more or less like a cat, a dog, playfull or not, etc. and, like it always happens, pleo will try to be everything to everyone, and inevitably will become a product for an average Joe, with average everything.... I wish they just already released the personality building tools and info to the community and let community take care of its needs; and focused on the undiluted, original creator's vision for pleo...
I really enjoyed Dogslife and Aibopet's personalities for AIBO. I hope pleo will get to that point soon.

[Aussie_Pleo]

> .....
However AIBO's exploration talents were eclipsed by a major and expensive design problem, "DHS" (Droopy head syndrome). After a short time with AIBO's head being stuck the right way (or think of it as the wrong way), it would wear out a slip-clutch in the servos. So most long term AIBO owners are cautious of leaving their pets alone, and careful when they hear "whirring" when the legs or neck become temporarily stuck. The Pleo design doesn't have this specific servo problem, but possible wear-and-tear is the downside to a more active exploring robot.....

DHS was limited to one particular model and there were many models.

I don't believe it would be fair to compare Pleo to an Aibo they are two different robots. Pleo is probably more comparable to an i-Cybie. But even the i-Cybie was more interactive than Pleo is in it's current state. Perhaps the full potential of Pleo has yet to be realised and hopefully continued Life OS updates will reveal that.

[andreivc]

Aussie,
I do support your criticism of pleo, primarily because I believe we should support the company, but keep things honest - we all (both customers and Ugobe) will benefit in the long term...
Now, having said that, I know that you are waiting for your Aibo to arrive. Aibo isn't that far ahead of the game, either, at least in terms of interaction with the user (after all, it's not that important how more advanced the technology is - it's all about the illusion of intelligence, right?). I suggest that you enjoy your 3-series first for a month or so, and then compare it to Pleo. Again, my point is AIBO isn't that far away from Pleo. In many respects, in fact, it feels more like a rigid, programmed robot toy than Pleo, which feels more like a living thing. I know, you are going to say that interaction with Pleo is nothing more than stimulation of three spots and canned reactions. Well, again, just wait and have your AIBO for a couple of weeks to a month, and see for yourself. I had all series of AIBO, except for the 1-st. I am curious to hear about your impressions.
A.

[Aussie_Pleo]

Aussie,
I do support your criticism of pleo, primarily because I believe we should support the company, but keep things honest - we all (both customers and Ugobe) will benefit in the long term...
Now, having said that, I know that you are waiting for your Aibo to arrive. Aibo isn't that far ahead of the game, either, at least in terms of interaction with the user (after all, it's not that important how more advanced the technology is - it's all about the illusion of intelligence, right?). I suggest that you enjoy your 3-series first for a month or so, and then compare it to Pleo. Again, my point is AIBO isn't that far away from Pleo. In many respects, in fact, it feels more like a rigid, programmed robot toy than Pleo, which feels more like a living thing. I know, you are going to say that interaction with Pleo is nothing more than stimulation of three spots and canned reactions. Well, again, just wait and have your AIBO for a couple of weeks to a month, and see for yourself. I had all series of AIBO, except for the 1-st. I am curious to hear about your impressions.
A.

Sorry if I'm offending anyone here guys, that's certainly not my intention. I don't know how many times I need to say that I actually do like Pleo and I do think it's an amazing thing. I have also said that Ugobe may realise it's full potential with further updates.

My comments are not about Aibo vs Pleo so I'm not sure where you get that from. I am making open observations as to what I think Pleo is now. Pleo is realistic no doubt about it, but at present it's just not interactive enough to keep it interesting. Perhaps if voice recognition were introduced (say respond to certain commands for example) that would certainly make it more interesting.

Thanks for pointing out that I am getting an Aibo (it's something I have always wanted, and I'm not sure why you feel it necessary to mention it), but I have never compared one to a Pleo and given the difference in price I think that would be silly in any event.

[junkroxy]

i think a comparizon is very difficoult, since the price of Aibo is higher thatn Pleo.And you know, sometimes the price makes the difference! :) I don't have an Aibo, but you know, Aibo and the cheapest I-Cybie are the first great consumers robot, before Pleo... I had I-Cybie, but i think he was a little boring... I'm glad Pleo it's totally different from them! :D

[sandro]

Perhaps if voice recognition were introduced (say respond to certain commands for example) that would certainly make it more interesting.

Or really annoying. As I was leaving home today, my Aibo mistook the creaking floorboards for someone talking (in an otherwise quiet house). Don't even get me started on what it's like trying to watch TV in the room with Aibo. Frankly if my Pleo starts having voice recognition, I'll probably end up leaving it off more often.


-Sandro

[andreivc]

Aussie, i mentioned you getting aibo in the context of how great an opportunity you will have to compare the two for yourself. Nothing else was implied, whatsoever.

Just a friendly reminder, that this thread was started specifically to compare the workings of the artificial intelligence in Aibo and in Pleo. As someone identified, pleo works very similar to AIBO in that department. Sandro, can you share your experiences with us? How's pleo's thinking diferent from AIBO?
Thanks,
A.

[PleoPet]

>... But even the i-Cybie was more interactive than Pleo is in it's current state....
FWIW iCybie is a great example on the importance of the software "personality"
The default iCybie personality had a lot of potential, but was underused IMHO and had a bad software balance (doing the same things over, not enough randomness). AFAIK most iCybie owners would agree that running an improved personality from a cartridge makes the iCybie a much better robot pet. Personalities like "ZCybie" or other similar personality upgrades were created by the user community.

The major difference is that iCybie needed a rare and expensive option (expansion cartridge and a way to program it). With Pleo the firmware update technology is built in (give or take an SD card or USB cable).

[Aussie_Pleo]

....The major difference is that iCybie needed a rare and expensive option (expansion cartridge and a way to program it). With Pleo the firmware update technology is built in (give or take an SD card or USB cable).

And like I said repeatedly here perhaps with further updates by Ugobe, Pleo will realise it's full potential.

But to compare Pleo in it's present state to an Aibo is not fair at all, Pleos are more comparable to an i-Cybie. That doesn't make either one a bad robot, it's just the way it is.

[andreivc]

Aussie, have you tried an AIBO?

[ryane]

in my opinion pleo and i-cybie are nothing alike. i own both. i also own an aibo 210. pleo is light years ahead of i-cybie in terms of interaction! i-cybie only has buttons to press, a "light sensor" (for petting), a so called motion sensor in it's nose for god knows what (i could never get a reaction), and of course ir for object detection. i-cybie is a very boring robot to interact with physically. it's entertaining to watch though.

pleo is amazing to witness first hand. i'm really fond of my aibo but pleo is something special entirely. it just makes so much more sense to pet and hold him to show approval instead of finding the correct head press or words to say.

[Aussie_Pleo]

Aussie, have you tried an AIBO?

Yes I just adopted one. It is more interactive and autonomous, but then for the price I'd expect it to be and that's why I think it's unfair to compare the two.

[PleoPet]

re: price.
It is also unfair because AIBO is old and the current eBay prices are much lower.
FWIW: A second hand AIBO (some around $500) is much closer in price to a brand new PLEO. The original iCybie price was closer to AIBO than most people think (iCybie originally $200 without accessories, second hand AIBOs were only 3X that). Because of market forces iCybie were commonly available for $50, which is the price that stick in people's minds.
Some additional discussion: http://www.aibo-life.org/forums/cgi-bin/ultimatebb.cgi?ubb=get_topic;f=1;t=006298;p=1#0000 07

--PleoPet (pimping for AiboPet)

[Aussie_Pleo]

Those prices you mention are rare as the thread you link it to mentions.

When Pleo can respond to it's name and 74 other voice commands, decide when he is bored to do something, play with an object (not just eat it), take pictures with his built in camera just as an example, then maybe it would be a little more interesting and more comparable to an aibo, that and the fact Pleo's battery is one hour play/4.5 hours recharge and Aibo's battery is 3 hours play/2 hours recharge. But it's features and value are still similar to that of a new i-cybie and comparing a new Pleo to the value of a second hand object is deceptive and moot.

So when I say you cannot compare the two, I'm not talking about price. I'm talking features.

[andreivc]

Aussie, I am enjoying the debate - what else is here to do as we are waiting for the OS upgrades...
This thread is about artificial intelligence;
having speech recognition and responding to 75 voice commands is not really a measure of AI; your cell phone can do that.
locating the object and kicking it is also not really the AI feature; you can download software from the net to have your webcam and PC recognize pictures; it's just a function.
taking pictures - the same thing; I don't see any AI in this.
and deciding when bored or not - well, pleo can do the same.

The decision making that the robot is doing (or capable of doing) in terms of what to do next, and why - this is where the AI comes in. If you read this and other threads and comments from people who wrote code for AIBO and other great robots, you will see that decision-making logic in AIBO and pleo are similar. As your AIBO roams around, you will see that it is as predictable as Pleo, but less life like, which is another issue.

[Aussie_Pleo]

The point remains aibos are capable of far more than a Pleo and for that I maintain the comparison is unfair. Decision making logic between Aibo and Pleo may be similar but that's probably because the limited amount of things Pleo can presently do might be compared to some of the abilities of an aibo and that's really where the comparison ends. Otherwise it's like saying your toyota has air conditioning, brakes, two headlights and a radio so therefore it is comparable to a Ferrari.

My point remains that Aibos have many more features and AI abilities than Pleo currently has and that makes it seem more interactive by a long mile. That's the whole point of an artificial life form.

I tried showing my Pleo my mother's canary the other day, it had no idea what was in front of it(was not capable of) even though Ugobe suggest this. The marketing of Pleo suggests it's ability to be more than what it actually is in it's current state and perhaps some people have gotten caught up in that marketing hype... I recall the same phenomenon with furbies, people were convinced they were more than what they were capable of doing also.

[andreivc]

I cannot see, Aussie, where that "mile" that you are talking about can fit in between Pleo and AIBO 3-series.
Kicking the red ball is pretty much the only thing that sets them apart. Well, that kicking get's boring very quickly.. after all, kicking is kicking, and kicking it is. aproach the ball... kick it... make a happy sound.... approach the ball.... kick it... make a happy sound....
and the learning is ... do the dance... be petted... do the SAME dance more often..... over and over again... Amazing!
plus... 3-s don't even look like anything that's a life. most of it's head is empty, did you know?
so that "mile" just ins't there.. more like a couple of feet; AIBO actually falls behind when it comes to looking like a living thing. But, I know--all arguements are kinda pointless since you JUST got one. Pleo is impressive too, in the first couple of charges, based on YOUR OWN comments about it in the appreciation forum when you just got one. You seemed to be pretty amazed with the technology. Both AIBO and Pleo are impressive at first. I wonder how long your AIBO fascination will last? hmmm....

[Aussie_Pleo]

LOL no need for the attitude andreivic.

Sure Pleo looks life like (but then you have chipping teeth, flaky skin because it's painted and the issues with it tearing as well), but that's the major appeal of it. You speak of kicking the red ball... how about 75 voices commands, recognising it's name and responding to it, media interaction (it responds to certain sound broadcast over media), 3 hours of battery life (recharge for two hours) etc, etc.

I guess by comparison you have stroke under the chin = sing, stroke top of head = purr, stoke back = coo (over and over again, amazing) and that's it...oh there's sleep also (need to conserve that 1 hour battery before it goes on charge for 4.5 hours). So yea there's at least a mile already :)

[andreivc]

yeah, I remember the imperfect voice recognition.... like someone said, I am glad pleo won't have that one... cuz there has to be perfect silence... and aibo must not move... for it to work.. ANd then "AIBO! Sit!"...one two three four five six seven eight seconds later - Aibo shakes head and makes a sound "I didn't get it!" You repeat again "Aibo! Sit!" one two three four .... and so on... but then it works sometimes! And it's amazing, I agree. But again, your cell phone has voice recognition too, and your computer operating system. For me, it was a torture, and I don't suffer from any speech defects :) One of the most useful commands were "go forward!" and turn around! and so on. ALmost never used any of them - just got tired of rediculus repeating, and there wasn't anything usefull anyway. Even take the picture wasn't that fun, after you see a couple of low-resolution aibo pictures, it's like "what's the point?" of it again. If you have seen the world of aibo through its eyes once, you don't want to do it again. Makes you feel sad for the robot.

pleo sounds drive me crazy, but AIBO sounds, especially the 3-series, just drove me nuts after a while. Endless chimes, no variety whatsoever. It had a volume button - very handy. FOr that reason, I sold my 3-series within the first month and bought a 210 supercore, which I sold as well, for all same reasons. and then got an ERS-7, the latest model, and that got sold eventually, too. I guess I suffer from that lack of imagination, I just cannot see AI where there isn't any ;)

[Aussie_Pleo]

andreivc,

I have no idea what you're talking about. My aibo responds on first command, even when the television is on, he can even hear me call him from the other room, you're grossly exaggerating or you speak in a very quite voice (have you tried asking your Pleo to do anything?) Perhaps your aibo chose to ignore you? After all that is part of it's ability as an interactive robot pet with AI, it adds a nuance to it's personality that is real and not imagined.

Too bad you didn't enjoy your aibo experience and learnt how to interact with it properly (as you mentioned you only had it a month so I doubt you even got passed the infant stage with it so you couldn't possibly know what it is capable of), but your comparisons are misleading for your belief of what Pleo can and cannot do (like I said people had the same idea about furbies and it's inventor admitted that furbies had no AI after the release of his new Pleo).

Incidentally what aibo forum are you a member of to tout here about my recent acquisition of a 311? Or would you prefer to keep that to yourself and keep one up on me? :rolleyes:

[andreivc]

Aussie, I had 210 supercore for over a year; I even traveled to another country with it; and have been involved with AIBO and AIBO forums since 2001. I have explored AIBO in depth; kept an eye on new developments - that's why I owned many throughout the years. i explored all there is to it on the average user level, including YAPT personality modifications, but short of serious programming - so i guess i did learn how to interact with AIBO :D and have a good undestanding of its capabilities, and can distinguish between AI, fantasy, and technical limitations of that product.
This is the Internet age; it's hard to keep secrets in our days, especially when you are chosing the same name in the aibo forum; I'll say hi to you when I see you in the aibo forum.

[Aussie_Pleo]

Maybe you forgot your last comment. Let me remind you..

....FOr that reason, I sold my 3-series within the first month and bought a 210 supercore, which I sold as well, for all same reasons. and then got an ERS-7, the latest model, and that got sold eventually, too.

A month isn't years and not really long enough to get to know the capabilities of an aibo. More exaggeration?

[PleoWorldHostRaine]

Just a gentle reminder.. It's wonderful to debate the issues and the pros and cons of things. But please remember to keep it a friendly debate. Let's keep our focus on issues, not personalities :)

[andreivc]

Aussie,
i thought i was clear in my post that i owned a 3-series for a month; but the next one i bought was a 210 supercore (because I thought the 3-series was too slow), and I owned a 210 for a year or so; then I sold my 210 to someone in France, and when a new model came out (7), I bought that one.

but really, Aussie, how long does it take to get to know the capabilities of AIBO? I guess it depends on a person.
Some people can figure it out right away and see how the cause and effect rules work in an AIBO, or a pleo, or any other robot. There are other people, that attribute certain events to some fuzzy logic or higher AI reasoning, or to a pure coincidence. But seriously, AIBO's are not that complex not to figure them out quickly. I bet you will figure out yours in a week. In any case, if you want to learn about how their personality works, you can always go to a resource like Aibopet's website: http://www.aibohack.com/. He gives you all the details about personalities, functions, and more. It's a great resource, including for those who want to get to know their AIBO like right away and not wait a month. :)

[Aussie_Pleo]

LOL yeah thanks for that, but I have the patience to work through my Aibo's natural progression :)

For the benefit of those reading this who don't own an Aibo let me fill in the blanks (I'm referring to a 3-series as that is what I own). Aibo Life is about your Aibo maturing into adult. There are several stages each with a sub-stage. They go from infant to child to juvenile to adult and each of those, like I said, have several stages of their own. The stages don't transition over an hour or so, in some cases they take days or longer depending how you interact with it. As the aibo develops through each stage (after infant) the aibo's personality is shaped by it's environment and this is explained further in the manual. So if you only had an aibo for a month it is most likely you never got to know it and realise it's actual potential.

Pleo on the other hand progresses from infant to juvenile in about 45 minutes and there is no further progression. That is it's so called "learning" phase. After that it performs routines based on certain patterns which are easy to recall (ie: scratch chin, pat head, stroke back, hold in arms) and that's it. Pleo presently cannot respond to any voice commands, cannot make a decision (Aibo will explore on it's own if it gets bored for example), it cannot develop a personality based on certain behavioural patterns (Aibos have stages within stages that are influenced by their environment).

There are other people, that attribute certain events to some fuzzy logic or higher AI reasoning, or to a pure coincidence and this is easily the case with Pleo because it looks so life like and has a baby animal way about it. Much the same as the furby, the Pleo is physically appealing and once again people are convinced it is capable of far more than it actually is. As I said countless times already maybe there's a potential there for a major OS update to make it more independent, but in it's current state Pleo is very basic and this is why I think it unfair to compare to an Aibo.

But seriously andreivc, Pleos are not that complex to figure them out quickly. I bet you will figure out yours in an hour. I guess that's why you prefer a Pleo? :)

[andreivc]

Aussie, like you said,
"A month isn't years and not really long enough to get to know the capabilities of an aibo."
and
"So if you only had an aibo for a month it is most likely you never got to know it and realize it's actual potential."

You have a lot to learn about Aibo, and maybe I can help. Here's an interesting tool you can use - or anyone else who doesn't want to wait a month for your AIBO to (painfully slowly) develop...
it's called ABrowser23 - among other things, allows you to set you aibo to any stage, including adult

http://www.aibohack.com/2or3/browser.htm

http://www.aibohack.com/2or3/ss_brow4.jpg

and you don't have to waste a month or more! very handy!
I am glad I could help
and I hope I alleviated any concerns of yours about me "exaggerating," and not having enough time to get to the adult stage with my 3xx :o try it, you will see they aren't exciting as adults either...

[Aussie_Pleo]

...I am glad I could help
and I hope I alleviated any concerns of yours about me "exaggerating," and not having enough time to get to the adult stage with my 3xx try it, you will see they aren't exciting as adults either...

You must have misread my post previous to this one. Like I said "thanks for that, but I have the patience to work through my Aibo's natural progression.... I guess that's why you prefer a Pleo"

I don't want to hack my Aibo's program as I am enjoying teaching it as it progresses through it's various life stages. That is after all the whole point of Aibo Life :rolleyes: (I wonder if you'd want the same thing for a real dog or would you just buy an adult one pre-trained?) If I wanted something that progressed in an hour and remained predictable for the rest of it's days we wouldn't be having this discussion.

Then of course if one did choose the course of hacking into Aibo's program there is still the benefit of a more advanced and exciting adult robot with a lot more capabilities... all you need to do is get to know them properly :)

[andreivc]

That aibo explorer reminds me of another unpleasant element of AIBO evolution....= trying to interact with AIBO in a 'right' way to keep the kind of adult that you want... that was like having a tamagochi=tedious...

For example, if you wanted to have a "Superior Adult" (a "balanced" personality that doesn't demand constant attention, but doesn't ignore your commands), but you forgot to "pet your aibo enough" or to "encourage it enough" to go after its ball, your AIBO's personality can "evolve" into a less desirable, lets say, Cutie, (see my post above) who doesn't want to do anything but sit and make cute motions. It was terrible! It was like having a tamagochi!

But thanks God, the mod community came up with brain surgery and the ability to freze AIBO's development. I guess many AIBO owners didn't want that kind of "learning" and "personality evolution" to take place... having experienced that anoying aspect of AIBO development myself, I totally understand it...

Rather than having AIBO adopt to you, it is YOU who have to adopt to AIBO, to make sure it doesn't change into a more boring robot. It's a full-time job.

[PleoPet]

Selective comments:
No need to fight.... However if you want to discuss the trials-and-tribulations of AIBO development, I recommend one of the AIBO boards, like aibo-life.org
---
IMHO: the original ERS-31x AiboLife development is tediously long. It can take over 3 months to get to an adult:
http://www.aibohack.com/310/evolve.htm
Also it looks like Sony realized their mistake and fixed it in the newer "Chatty Life" for the 31x which has a quicker evolution, and a user settable "speed up" option. Closer to a month and a half, similar to the 2x0. For the ERS-7 models they went too far IMHO (very fast and simplistic 'evolution')

Also true that AIBO evolution (for all models) is much much slower than PLEO. Even in the best situations, it takes ~6 weeks of active play for an 2x0/31x AIBO to grow up to an adult. That's why I wrote various "Brain Surgery" tools to speed up the process, and optionally freeze some personality changes. Whether it grows up naturally or with "Brain Surgery" is up to the AIBO owner to decide.

Bringing this back to PLEO - this is yet another example of "personality balance". The software writers are only guessing at how the personality will be used. Whenever they pick a length of time for each "stage" of evolution, some owners will think it is too long, others will think it too short. When the PLEO personality becomes more complicated, "Brain Surgery" tools will likely be available for PLEO as well.

[Aussie_Pleo]

....if you wanted to have a "Superior Adult" (a "balanced" personality that doesn't demand constant attention, but doesn't ignore your commands), but you forgot to "pet your aibo enough" or to "encourage it enough" to go after its ball, your AIBO's personality can "evolve" into a less desirable, lets say, Cutie, (see my post above) who doesn't want to do anything but sit and make cute motions....


In other words Aibo's develop according to their experience and environment, which is what we were discussing in this thread when the comparison between the two came up. You have described exactly my point which I have been making all along.... thanks for that :D

Pleos do not have the same ability, within an hour they "evolve" from hatchling to adult and the result is the same with the same predictable interaction for all of them. Aibos, as you have explained so clearly, develop based on human interaction.

[andreivc]

Aussie,
Please, read my previous post one more time, but with selective perception switched off. :D

[andreivc]

Also, please read Pleopet's (Aibopet's) post. He is the author of www.aibohack.com, The Authority in the world of AIBO.

[andreivc]

And nobody is saying that Pleo's evolution is more complex than AIBO's.
But, AIBO's evolution is long and annoying, in my opinion. I hope Pleo is not going to turn into anything like that with future OS upgrades.

[Aussie_Pleo]

Aussie,
Please, read my previous post one more time, but with selective perception switched off. :D

Okay..... and thanks for agreeing with what I had been saying all along :D

[pirpintine]

now now guys, no need to fight. i own and aibo and a pleo and i like them for very different reasons.

true, aibos AI is much more evolved than that of pleos. at the moment. but pleos still getting updates and improvments. im sure it can catch up to a degree at a later date, at least i hope so. haha.

i thought the growing up for my aibo did take a whiiiile. but it probably wanst overly helped by the old battery i had at the time which lasted about 40 minutes. if i was lucky. but i didnt find the lenghthy process boring at all. because it was quite sweet to watch him grow up and do new things at his own pace.

im hopeful for pleo at the moment. im sure he will grow up AI-wise too. but at the moent i think aibo is far more superior. but i love them both.