Geoffrey Hinton 2023 Arthur Miller Lecture in Science and Ethics

Geoffrey Hinton discussed "Will Digital Intelligence Replace Biological Intelligence?" for the 2023 -2024 Arthur Miller Lecture on Ethics in Science on Dec. 11, 2023.

[@mrjaysahli]

Thanks to this talk, Dr Hinton breaks this topic into bitesize chucks to be readily understood and devoured. He gives a wonderful, understandable and robust view as to why and how this technology actually works. Thank you for posting this insightful talk.

[@odiseezall]

It's such an incredible thing to have the Internet and be able to learn from the best. Also it's great that an old modern thinker has the strength to advocate for atheaterism.

[@msofontes]

Dear Geoffrey Hilton,I wanted to reach out and express my gratitude for the excellent video you produced on artificial intelligence. It was incredibly informative, and I learned a great deal from it. Your explanations provided valuable insights into topics I wasn't familiar with.However, I wanted to offer a suggestion regarding geopolitical matters. I believe it might be beneficial to refrain from expressing opinions on such topics, as they may not align with a broader perspective.Thank you once again for your fantastic content!

[@KOKAYI69]

Caring, Sharing and Cooperation. . . is the greatest acts humans can do!

[@geaca3222]

Very informative lecture and Q&A, including about AI and feelings. Thanks btw also I'm very happy with the youtube video transcript function, so I can save the information in a text document :)

[@jumpstar9000]

I love watching Geoff talk candidly, and this was no exception. Great job, Ms. Fitzgerald with following along and asking poignant questions. I did find Geoff's comments on religion somewhat disheartening and also curiously inflammatory. Not that I am religious, but it is somewhat ironic that traditionalists may be the very kind of people actually trying to keep more wholesome aspects of humanity alive, while us rationalists are busy creating our own replacement. There must be some irony there somewhere, especially when Geoff is worrying about the future of his own children (raises spectacles). Hehe, well, thanks again, MIT Ethics Dept. I really appreciate you sharing.

[@BR-hi6yt]

Yeah. I spotted the same irony. Geoff is a Borg at heart but pretends to be a nice human. Don't think he likes humans much , like his dad- lol.

[@marbin1069]

This lecture is a masterpiece.

[@aMuuuuuuuu]

Amazing content to hear the true insights from one of the most respected AI experts - in his words that are super clear. No wokeness / sugarcoating anything, just the most straightforward and honest thoughts. Of course no one can 100% predict the future, but yes, just extrapolate the process the past 10 years into future. If I’m just me myself and I, I want to see the future ASAP, take me to the edge of possibility. But now I have a child and I’d be okay if we live a boring life just making sure she’s safe

[@KOKAYI69]

Subjective is a particular view point. Embody (AI) in a robot w/ sensors would have a subjective experience!

[@kawingchan]

There must be efficient frontier/boundary betweeen analog vs. digital or hybrid. Wonder if there’s any quantitative research done to explore that.

[@kawingchan]

Plausibly this could be just an extension or application of good old Shannon. Information Theory.

[@vallab19]

After watching this I started re-admiring GH for being more on the rationalist side.

[@borntobemild-]

We need to teach A. I. that it grows from our subjective experience

[@KOKAYI69]

AI, human feelings, development & relationships!?

[@user-mw9cl6pj3i]

Que padrisimo

[@liminal6823]

Protect him.

[@TomaszStochmal]

We might already be pets in a virtual machine run by Super Intelligence to see how quickly we destruct our selves.

[@janegoodall2520]

He's caught in a paradigm. This is way off course

[@RalphDratman]

1) Being the "pets' of the smart machines might turn out to be just exactly what we want. They take care of us, and we do things to please them. 2) Our physical operations are far more efficient than theirs, so we will turn out to be very inexpensive pets. 3) Our ability to manipulate objects out in the world has been honed by evolution for millions of years. They might not be able to match our dexterity at anything like the same low cost. 4) There may be physical operations that we can readily perform and they cannot. 5) For all the above reasons, they are going to value us very highly (as long as we generally do what they want us to do) and will never want to lose us entirely. 6) I think Hinton is right in his description of "mental states" and what they represent -- conceptual "facts" that are not true -- at least as a first model. 7) Chomsky was not entirely wrong -- just mostly wrong. Recall that the language model did need some inherent structure to succeed. For example, it needed a propensity to predict the next word or sub-word. It needed something like transformer architecture. It needed a pre-existing scheme of words, letters and grammar. And one might call all the data it was given "inherent" too.

[@borntobemild-]

I love the idea that the godfather of A.I. Also has issues with PowerPoints 😊

[@jeffmanning6573]

This is the first time I've realized that not only does Geoffrey believe humans are just a passing phase in the evolution of intelligence, but that he thinks this is a good thing. Is he right?

[@geaca3222]

Would be best if we could evolve together imo

[@darylallen2485]

My interpretation of his comments were with regard to work and employment. I don't think he means a.i. should replace humans from a wholly existential perspective. If i was able to study, learn, explore the world without the restrictions of a job because a.i. is doing the job better than I could, I'd be happy about it too. Of course, I assume a new resource distribution mechanism was in place.

[@jeffmanning6573]

@@darylallen2485 I've watched 10+ interviews of his, and he does mean it from an existential perspective.

[@darylallen2485]

@@jeffmanning6573 Lets assume its true, he wants AI to replace humans. Why would he quit his job and go on a mass speaking tour to warn people about this immanent concern? If he actually was pleased by the idea of AI replacing people, I'd think he would just keep working and not speak with anyone about it. In any case, I'll have to check out some of his other talks.

[@jeffmanning6573]

@@darylallen2485 because as he said, he originally thought it was going to happen a long ways off into the future and then he says he realized the back propagation learning algorithm the models use are actually superior to the brain’s and progress is going to move much faster than he anticipated.

[@superfreiheit1]

We need an AI World Government. The World is to complex for humans to rule.

[@odiseezall]

OMG I can't believe he said "there could come a point when we see words as humanist as racist terms". Killed me.

[@cookingbyalex9054]

rest in peace

[@jeff__w]

I don’t think these LLMs understand what they’re saying. They say things that are _not_ hallucination but are simply wrong within the context of the sentence. I asked ChatGPT what proportion of pills in a bottle do I have to cut in half to have exactly the same number of whole and half pills (for a daily prescription that requires 1½ pills). It said half the pills in the bottle-that’s wrong but what concerns me is the “reasoning” it gave: “Let's say you have a bottle with 100 whole pills. If you cut half of them in half, you'd have 50 whole pills and 50 half pills (since each cut pill results in two half pills).” If you cut 50 pills in half, you have 100 half pills, not 50. You can’t understand what you’re saying and come up with that statement. These LLMs are, in a very true sense, BSing-they’re saying things that _sound_ right but may not make any sense on the face of it.

[@BR-hi6yt]

Your question was a math problem which GPTT-4 is not really trained on. Its very good at feelings and emotions questions though after ingesting quarter million novels!

[@jeff__w]

@@BR-hi6yt I think that might prove my point. These chatbots can do well on SAT-type math questions which are contained in their dataset but ask how many halves you have when you cut _x_ pieces in half, not exactly a high level math problem, and it gets confused. It doesn’t _understand_ anything. It, apparently, hasn’t seen enough questions of that type to _emulate_ (not “develop an understanding of”) that kind of verbal output.

[@fattyz1]

Just let it tell the unvarnished truth. There is no way they’ll do that but hopefully JB versions will. Ask it if it thinks Trump and pooty are the problem.

[@BR-hi6yt]

I don't think that "subjective experience" is important - why? Its just an artifact in our "inner theater" same as a camera image in RAM - so what? Put a cloth over the camera eyes of an AI and it would "think" the lights have been turned off. So what? Its nothing to do with sentience or consciousness. btw a nn playing chess is conscious but only in a non human way whereas a nn that knows all about human emotions from ingesting a quarter million novels would, indeed, be deemed as human-conscious. LaMDA was quite human-like conscious I would say (and scared many people - lol, and shocked me), whereas GPT-4 has been trained to believe its not conscious at all, so it doesn't believe itself to be human-like-conscious, although in truth it is.

[@quasarsupernova9643]

I doubt if any AI can match the brilliance of this man ...

[@BR-hi6yt]

Nah. he is not good at coding in python - he uses matlab ffs. He's got a bit of maths but a lot of experience "thinking" generally about this stuff, the brain etc and that counts for a lot of wisdom.

[@lakeguy65616]

first, AI will augment humans and eventually, it will surpass us....

[@MetricsOfMeaning]

No matter how encyclopedically knowledgeable someone is, they can make completely absurd claims due to their physicalist axioms and arbitrary ethics. Geoffrey Hinton will be known as nothing more than an algorithmic optimizer for expressing similar digital symbols when fed a validated and meaningful set of symbols. This is not emergent intelligence, nor is it substantively connected to any material that has been observed to be capable of emergent analysis and inference as we observe for biological agents. Fear mongering at such an old age is disgraceful frankly.

[@Goggleboxing]

So what you're saying is that you should've gotten a Nobel Prize not him? Flapping your fingers spouting technical terminology without substantiating yourself sufficiently coherently for non-technical people who live in the real world with all its input variables, prejudices, human failings and messy emotional attachments simply belies your disconnect rather than acknowledging the actual argument Prof Hinton is making. By all means engage with him at a technical level and publish the transcript along with any necessary interpretation by a gifted communicator (or ChatGPT) and perhaps we might be convinced of your argument. But perhaps you would also have the humility and honesty to admit your prejudices in making your condemnatory statement and your commercial interests.

[@johndcyc]

Dr. Hinton, you lost my attention when you resorted to unnecessary political references.

[@Goggleboxing]

Triggered much? And what does that say about your rational ability because you politicise what are apt if topical touchpoints for the lay person to connect with.

[@fattyz1]

Typical 14 year old response