Apple Is Working On LLM-Powered Robots, Report Says (bloomberg.com) 42
Apple is secretly developing robotic devices powered by generative AI, including a table-top robotic arm with an iPad-like display and a mobile robot for household chores, Bloomberg News is reporting, citing people familiar with the matter.
Generative AI LLM (Score:2)
An LLM is just one type of AI system. It wouldn't make sense to use it for something that requires enviroment comprehension and motion unless the definition of LLM has been stretched so far as to be meaningless.
Re: (Score:3, Informative)
Re: (Score:3, Informative)
Your presumption is correct, the article is MOSTLY 22 out of 24 paragraphs) about how OpenAI can help improve Siri.
Re: (Score:2)
Web sites behave differently depending on if you visited the site before, geolocation and the IP you use, if you browse in anonymous mode, the browser and OS you use (user agent), if you are logged in into your Google account and the likes, etc. etc.
Everybody gets his own view of the Internet nowadays... /s
Re: (Score:2)
Re: (Score:2)
Because you read less news and aren't hitting the free paywall limits?
Re: (Score:2)
Re: (Score:1)
archive.is is your friend in such cases.
But to quote directly from the article:
Apple also is looking beyond chatbots. It aims to use large language models — a key technology behind generative AI — to help power a pair of robotic devices that it is secretly developing, the people said.
That includes a table-top robotic arm with a large, iPad-like display. The company also has been working on a mobile robot that can follow users around and handle chores on their behalf. And it’s looking to equip its AirPods with cameras and AI features.
Not a lot details, could also just be technical experiments.
Re: (Score:2)
Traditional tech companies are desperate to find the next hype to raise their profit margins furthermore. The truth is that we have reached a plateau where we already have everything we need technology wise. Some even make matter worse by somehow trying to innovate, take Google search or Microsoft recall as an example. I haven't spend a dime on technology for years excepted replacement parts like failed hard drives etc.
Re: (Score:3)
we have reached a plateau where we already have everything we need technology wise.
People have been saying that for centuries.
Re: (Score:3)
Yes, mostly the people who say it are older. That's also been the way it's been for centuries.
Re: (Score:2)
“I've come up with a set of rules that describe our reactions to technologies:
1. Anything that is in the world when you’re born is normal and ordinary and is just a natural part of the way the world works.:
2. Anything that's invented between when you’re fifteen and thirty-five is new and exciting and revolutionary and you can probably get a career in it.:
3. Anything invented after you're thirty-five is against the natural order of things.”:
Douglas Adams, The Salmon of Doubt: Hitchhiking the Galaxy One Last Time
Re: (Score:3, Funny)
“I've come up with a set of rules that describe our reactions to technologies:
1. Anything that is in the world when you’re born is normal and ordinary and is just a natural part of the way the world works.
2. Anything that's invented between when you’re between fifteen and thirty-five is new and exciting and revolutionary and you can probably get a career in it.
3. Anything invented after you're thirty-five is against the natural order of things.”
(Douglas Adams)
Re: (Score:2)
I'm sure Apple (or anybody for that matter) would love to create a chores robot, and the quote is probably a reflection of that. (Maybe Apple thinks that prompt based assembly line arms leads in that direction?) It's not happening anytime soon though.
Re: (Score:3)
The choice of name for "LLMs" is terrible. The models themselves are generalists, and happily handle any kind of data from any modalities, including mixed (across modalities**). It's just that language processing was the first thing they were applied to. It would be best if people simply referred to them as their underlying technology, transformers.
A good article on the topic [arxiv.org]:
Re: (Score:2, Interesting)
No, the name of LLMs describes exactly what they do. They are language models. Slashdot's penchant for conflating AI, LLMs, generative models, deep learning, and etc. is the problem.
Artificial neural networks are generalists. Famously so. You can use a particular arrangement of an ANN called a transformer to mix data in such a way that is quite useful for language models. You can also use transformers for other things, but there are generally better ways.
Re: (Score:3)
Transformers works just as well on non-linguistic tasks. It is absolutely not just for "language models". You can plug Transformers into protein folding, or weather prediction, or CFD simulation, or basically anything, and they perform superbly across all of them. It has nothing to do with "language".
When applied to language, the first thing you do to Transformers is get rid of everything having to do with language. First you tokenize, so it no longer has anything to do with words or sentences. Then y
Re: (Score:2)
The comments section of this article is full of people being confused because the article says "large language model" and the task isn't language. Whereas nobody would have been confused if we had just from the start referred to the technology by its name, transformers. But now people have stuck in their head that this technology is all about language, so what could it possibly have to do with a robot? When you can have any mix of modalities in the mix, and they all interact cooperatively and benefit each
Re: (Score:2)
I didn't say transformers work only on language. They work well on one dimensional signals where there are important long-range correlations, like language. Yes, they also work well for reading DNA and RNA, which really are a kind of language.
They are much less efficient for things like images and sound. They work fine of course, but usually need very large models and a lot of training data before they approach the performance of other techniques.
If you are getting excited about something I didn't say becau
Re: (Score:2)
This is simply false - they work great with multidimensional positional encoding. Tokens are always outputed in a sequence, but the positional encoding can have any number of dimensions.
People make this sort of stupid analogy a lot (as if it's some sort of a truism), but they really aren't except at the most vague levels. Protein folding involves 3d physics, just from
Re: (Score:2)
Try Be my Eyes if you have a moment. LLMs can be very useful for environment comprehension. Many people seem to think that AI ist just good for chatbots, but this really is just the tip of the iceberg.
https://www.bemyeyes.com/blog/... [bemyeyes.com]
risk centric bizdev (Score:1)
misleading summary (Score:2)
Also, how much of a secret are they if Bloomberg is reporting on them?
Re: (Score:2)
Also, how much of a secret are they if Bloomberg is reporting on them?
Secret enough that they had to pad out 2 paragraphs into 24 paragraphs with meaningless speculation.
Re: (Score:2)
and you doubted LLM capabilities
T-001 is a good name... (Score:2)
T-001 is a good name for autonomous AI killer robot...
Re: (Score:2)
T-007 will have a medicine applicator. Bond. Gold Bond.
LLM is too dangerous (Score:2)
We're still at a point where we can convince an LLM that "kill all humans" is a reasonable way to interpret Asimov's 3 laws. Giving it any agency at all is just a human hallucinating that an LLM has intelligence.
Yes! (Score:3)
How good are their robots? (Score:1)
If they still need near-slave labor to assemble iPhones how can their robots be any good? If they can't make a robot that can put together an iPhone their home robot must suck.
LLM stands for Large Language Model, right? (Score:3, Insightful)
This headline is fucking incomprehensible. "Powered" by LLM? Aren't robots usually "powered" by a battery? WTF is a robot going to do with an LLM anyway? You could put a speaker on a stick and it could talk several languages at you in statistically generated sentences, but robots need neural processing as far as I know, especially for visual interpretation. The article is paywalled, has a completely different headline, and TFS is woefully inadequate so I can't even read that and understand. The only thing the headlines have in common is "Apple," and they both appear to be a cry for relevancy when they missed the fucking boat because Jobs died.
Maybe we're talking about a natural language interface so it can do complex tasks ordered by confused, inarticulate consumers?
Customer: "Clean that shit up!" Robot: "Sí señor. Limpiando tu culo!" Customer: "Ow, ow, ow, OW OW Wheeeelahhh."
And whom exactly are we citing? Are they Apple middle management, engineers, C-suite? What qualifications do they have? Tech? Marketing? Janitorial? Sartorial? Editorial?
WTAF? How did this even get published here?
Re: (Score:2)
Household robots? Good luck with that! (Score:2)
My robot floor cleaner must have been programmed to find new and stupid ways to get stuck in places it has already seen dozens of times.
My robot pool floor/wall cleaner can sometimes make it to the second step. I usually run it 3 times in a row because it can't figure out how to navigate my industry standard rectangular pool without missing spots.
My robot pool skimmer seems designed to dodge floating debris and hit the walls as hard as possible, pause, back up, hit the wall again just to make sure that was
Re: (Score:2)
Robot will add extra thumbs to all human hands (Score:2)
A hand has five fingers, some people don't consider the thumb a finger, and two thumbs up is a common phrase, as is thumbs down, so it should have at least eight finger-like digits on each hand, right? Stay real still and let me sew these extra fingers on to your hand, human. Try not to question how they were acquired. That isn't important right now, human.
Apple robotic arm? Just have to wait a bit. (Score:2)
I'll be certain to go purchase one once I have an Apple Car to go pick it up with.
I am sure Apple has plenty of R&D project ongoing. That does not imply we will get any products directly from this R&D. It is just a test to answer the question "Can we do it?". If yes they still have to confirm that there is a market, it is economically viable, and that there is no better use of their time. It is way too early to speculate on this being an Apple product.
Pretty neat idea (Score:2)