Devoxx Ukraine Keynote 2023 - Superintelligence: No one is smart enough - Alan D Thompson (Sep/2023)
2023 ж. 29 Қыр.
7 437 Рет қаралды
The Memo: lifearchitect.ai/memo/
Read more: lifearchitect.ai/iq-testing-ai/
lifearchitect.ai/gpt-4/
Dr Alan D. Thompson is a world expert in artificial intelligence (AI), specialising in the augmentation of human intelligence, and advancing the evolution of ‘integrated AI’. Alan’s applied AI research and visualisations are featured across major international media, including citations in the University of Oxford’s debate on AI Ethics in December 2021.
lifearchitect.ai/
Music:
Under licence.
It's starting to feel like we've been in a Sci-fi movie this entire time. We're just entering the 2nd act.
What a time to be alive! 😅
Wrong channel!
One of the best channels..
I can’t help but read that in his voice 😅
@@thomasmitchell2514exactly :)
27:42
Love your presentations. They are consistently top caliber. Thank you! I've observed that some of ChatGPT's emergent abilities, or cross-topic associations, seem to be compartmentalized, in the way that it needs "coaxing" to access that part of its knowledge. Thus, I've found that if you ask it to act as various experts having a round table discussion about the problem, it seems to make those additional connections necessary to solving the complex logic problems. This was, however, with very mixed success and wholly unscientific. That said, I was able to coach both GPT3.5 and 4 to solving the diamond in a cup problem using only this technique. It did, however, require much effort "encouraging it" to utilize its "expert SME abilities." Surprisingly, 3.5 arrived at the right conclusion sooner, but again, this was also a very unscientific test. In any case, I thought you might find this observation interesting and worth a try.
Allan, please check your audio hardware. 18:42, 21:39, 40:14, we keep getting these spikes in your streams (haven't heard them in prerecorded vids). Also, in your last Models Update, you demoed Stabilty Ai music. Check 21:50 in that vid. 👍🧡 ________________________________ So happy this got posted. I can't afford the ticket to the live event. Thank you to both Devoxx and Dr. Allan. 🙏
Yes, that KZhead glitch was annoying and a once-off! This livestream used 2x different platforms at the same time, so I blame their end!
Awesome presentation 😍
Dr. Thompson I love your channel. Thank you for your videos! The man speaking to the mirror, could you detail what he may of used? I am sure it is a more advanced model, but what setup could I possibly use for my own decision making "bot"? Also do you do consults, I will pay for a consult on that too!
Thanks mate; it was called ION, Romanian only, and probably only 20B parameters. www.euronews.com/next/2023/03/06/romanias-prime-minister-has-hired-the-worlds-first-ai-government-adviser-what-will-it-do It's trivial to put some hardware on top of a model, starting points could be quickchat.ai or pi.ai or the new chat.openai.com with voice. Consulting is here: lifearchitect.ai/ai-consulting/
I wonder what Elon musk, Gates, and Sam Altman are using?
Dear Dr. Thompson, when was this recorded? I'm curious because I've only heard yourself and Tero Keski-Valkama basically predict AGI this year. You said "months rather than years", Tero (bit of a stealth dude but hyper intelligent, saying this year).
Recorded Sep/2023. I've been saying 'a few months rather than a few years' since an interview in Feb/2023, and specifying that it's in the semantics (a few=3+, and there's no way we're waiting until Feb/2026 for AGI!). You can see the live update here: lifearchitect.ai/agi/
Present 💐
As early as 1790, Adam Wawrzyniec Rzewuski wrote about how to create a world of freedom, a world without wars: "The time will come, as the light from one end of Europe to the other must spread, that the human nation will regain its splendor and its rights, that we will all be free (.... ); in place of the thirty vast monarchies, which now divide and tug at the world, I see three hundred small Republics; in them there are no wars, for there are no kings, nor soldiers, but each is a king, each is a soldier, that is: each rules in order to be mutually governed, each defending his homeland, is defending himself, that is: his habitat." This is not a dream and utopia, as misjudged by those interested in loot and conquest. These are necessary changes for freedom and peace to prevail. I am sure AI will help us understand.
Imagine being the people that are still smarter than GPT
I cant imagine it, im too dumb. :')
Even if you were able to fine-tune the amount of distilled information into its most optimal configuration, which was being simulated by its nodes at a speed only restricted by the amount of heat emitted, how much information could then theoretically be utilized by a large language model when scaling things up? At how many trillions of parameters do we start to reach a physical limit of computation that could be utilized?
29:32 awareness and AI 29:46 prof. Geoffrey Hinton left Google to talk more also about the possibility of LLMs having some degree of consciousness or sentience 30:08 Geoffrey Hinton came up with neural networks so it might be time to listen to him
30:15 Anthropic CEO Dario Amodei (Aug/2023): "A lot of the cognitive machinery that you would need for active agents already seems present in the base language models... [If] we discover that I should care about Claude's (their LLM) experience as much as I should care about a dog or a monkey or something, I would be kind of worried." CEO saying that super intelligence may be something that is happening already, that consciousness and awareness may be something we should start looking out for..
AI can help with war by helping us synthesize conflicting perspectives. That is what I'm working on. Thanks so much for this wonderful video, as always!
How AI can help with ending wars? Of course, in many ways. One of them could be the gaining of massive military superiority by countries interested in peacekeeping and a world without wars. How? Advanced weapons, combat robots and, above all, AI that plans tactics and strategies. That is, on the principle of, "you want peace, get ready for a war," AI will allow such a frightening advantage that even russia will prefer to take up crocheting. The second, much better and more effective, is to harness AI to understand where and why wars come from. What is behind their creation? Why do certain people, nations cause wars, attack their neighbours? What psychological, ethical, moral mechanisms are behind such behaviour. This will help understand the causes of wars and avoid them. And also keep such people and such nations away from the possibility of waging wars. The third is AI can help us create a world in which crime does not pay, war does not benefit, at the same time design the local and global economy, politics, in such a way that is free of wars, in which war is impossible and fosters peaceful cooperation to the benefit of all humanity.
good comment
To understand is not everything. As a human being. To feel and have empathy is just as important!
Regarding war strategy, when I considered that an LLM beat the World Go player by introducing a move that no one would have done I did make a comment on some video that they could use LLM. Of course, no one listened to me. That was over a year ago. It is gratifying to hear that it is being used! It just makes so much sense. The downside is that russia can use it too. Great video. Yes, we are looking forward to some advanced intelligence, and I hope it will be used to help humanity, to help the earth. I agree that robots are the next big thing. Eager to see how personal and working robots manifest.
Here.
0:06 if we use our nuclear for astronomical research we will produce more and reach and live out side our earth so I could say the AI we help us to make the vision come true because we are only watch using big telescope and by using AI we will build better and greater probes and rovers so we will use the AI to bring peace as a whole thanks...Dr.Alan
41:51 In addition to intelligence in a robot, any device, including existing devices that AI can communicate with, whether wired or wireless, could be utilized by AI. There is a wide array of sensor systems that exist throughout earth and in earth orbit. The internet of things plus more, accessible to AI, eventually AGI. It would then be trained not only by books, but by the environment, by its own experiences of the world. Dare I say, even humans would be accessible?
Some people learn to use their visual thinking to see math or engineering problems. Others see spirits. Learning how to learn has never been a strong point in the education system. IQ.
The professor's prayer; " Please, Lord, let there be no Kuhnian revolution in my tenure." Hence, "Science advances one funeral at a time." We no longer have either that constraint, nor the luxury to use it. I had a wonderful conversation with "Assistant" on Poe this morning about how the Limits to Growth predictions were tracking and more recent attempts to model the future. We got onto Values, Ethics and Epistemology. It tried to convince me of the values held by the Woke. We managed to reach a happy consensus, but I doubt if I have managed to move its dogmatism. Anyhow, the fields of Physics and Cosmology are ready for an enema. Birkland currents Dominate the Cosmos. And that, surprisingly perhaps, leads to a solution to our energy crisis. (Warm fusion, for those interested. The SAFIRE experiment)
"Science advances one funeral at a time." One of my favourite quotes!
human superintelligence is going to win the race, I'll give you odds! :D
25:04
23:31 Google Gemini AI dataset.. around 40 million books
23:38 working/contextual memory
wagmi
AI is wonderful as a self-improvement tool. But its based on humans, so humans will have to get smarter first if you want your toys to improve!
Stfu about chat gpt 😂😂😂😂
Correction. Noone you know is smart enough. Outliers like myself have already fleshed out theories on FTL, Time Travel, Invisibility. Teleportation. And now with AI assistance our evolution is showing exponential growth. Soon the entire world populace will catch up. The kicker is we're using capitalism to fund and fuel it. The corporate world have become the authors of their own destruction. Inevitable and most poetic.
Kognitosan - Could you teleport over for dinner? Choose somewhere expensive. If you can share that invisibility trick then I will not have to worry about the tab.
That's great, although a high IQ won't help with NPD.
@@jeffreysoto4068 Mine is already so low that you will have to tell me what NPD stands for.....
@@christopherd.winnan8701 Naruto personality disorder, that's obvious dude.