This is a very educative presentation on how amazing technologies like AI can be harnessed to make us Awesome Humans. The new future that is here needs the much-needed education reboot, to make a paradigm shift to the Big Purple. To get the monkey off our backs, by pushing the monkey work to AI, we need to lose our routines to AI as we embrace higher purpose(Humanity, Ethics, Creativity, and Imagination) in Life and learn to partner with Intelligent Assistants. Together the planet, people, prosperity, and purpose are all at a tipping point that human history may have not experienced before.
Introspectively AGI sounds like a science fiction concept rather than something academical, you can't compete biological intelligence that has shaped and adapted for millions of years with a binary mathematical system, it's nearly impossible that a 100T+ parameter model would beat human at everything, I must admit it's easy to be fooled by AI thinking it's truly intelligent when GPT-4 sounds much smarter than you are where actually it was trained to predict approximately the plausible answer (fitting 2000 words maximum) for an input by learning billions of patterns between millions of books and web data, it's basically a text generator even though it can learn new modalities by labelling them with text, but I truly doubt GPT-7 would be able to write a (consistent) 300 pages thesis about topological insulators backed by actual experiments' reports conducted on superconductors in cutting edge labs instead of some hypothetical claims coming from hallucinations, AI will eventually get smarter and perform better than humans in most of the mathematically deterministic fields such as mathematics, physics and some sub fields of biology but it's probabilistically impossible that a text generator with no abstract thinking dimension or systematic intuitive introspections would be better than human cognition fundamentally. I think people referring to AGI actually refer to a 100T parameter model, a gigantic AI model that should technically learn the most subtle and unapparent patterns of its pretraining dataset which even the world class human experts haven't discovered yet.
Me gustaría más bien que usará la inteligencia artificial para ayudar a pagar los incendios más rápidos para localizar a personas que se encuentren en peligro para que se acaben las guerras para hacer tecnologías de siembras y cosas que en vez de quitarlas buscar forma de sembrar y de que los cultivos sean más eficaces y no tener tantos aparatos y tecnologías como para disfrutar mientras el mundo se cae casi quiero que vuelva la era de piedra para que se acabe todo esto
1 trillion IQ in mathematically deterministic fields only, because no matter the number of parameters there's limits to what the model can learn from its pretraining dataset when the knowledges in that dataset are mostly linear, lately synthetic data have been useful for mathematically deterministic domains such as math, when these models are backed by symbolic engines they can generate high quality synthetic data.
@@TheRealUsername cognition math doesnt even add up like that; you cant take 5 people with 100iq and expect them to perform like a 500 iq. also talking psychometrics, as a science today doesnt even work with those concept, what do we have as an absolute celing for todays metrics? 220iq Terrence Tao? trying to referce orders or magnitude with these metrics parameters its just not really something i can imagine.
5- AI, in order to improve its performance and prevent undesirable consequences, must continuously interact with “effective rules and stable principles in the realm of existence”. @jamshidi_rahim
The best AI talk I have seen so far!
Thanks !!
Excellent presentation. Unique and holistic. Thank you Gerd Leonhard. . 🙏
🙏🙏
Gerd Leonhard... that was a digital enlightenment. I think more of this presentation can save the world 🌍
Glad to hear that
48:32 wow the hubris. Silicone hustler. " I'm sorry Dave. I'm afraid I can't do that."
Thank you for your content
This is a very educative presentation on how amazing technologies like AI can be harnessed to make us Awesome Humans. The new future that is here needs the much-needed education reboot, to make a paradigm shift to the Big Purple. To get the monkey off our backs, by pushing the monkey work to AI, we need to lose our routines to AI as we embrace higher purpose(Humanity, Ethics, Creativity, and Imagination) in Life and learn to partner with Intelligent Assistants. Together the planet, people, prosperity, and purpose are all at a tipping point that human history may have not experienced before.
So true we don't Robots do to everything they can do just a few things but still make us quit
❤❤❤❤❤
A-we-sum 😊
Permaculture is good.🌎🙂❤️
ok but even 20% unemployment is a disaster.
well wait for AGI
Introspectively AGI sounds like a science fiction concept rather than something academical, you can't compete biological intelligence that has shaped and adapted for millions of years with a binary mathematical system, it's nearly impossible that a 100T+ parameter model would beat human at everything, I must admit it's easy to be fooled by AI thinking it's truly intelligent when GPT-4 sounds much smarter than you are where actually it was trained to predict approximately the plausible answer (fitting 2000 words maximum) for an input by learning billions of patterns between millions of books and web data, it's basically a text generator even though it can learn new modalities by labelling them with text, but I truly doubt GPT-7 would be able to write a (consistent) 300 pages thesis about topological insulators backed by actual experiments' reports conducted on superconductors in cutting edge labs instead of some hypothetical claims coming from hallucinations, AI will eventually get smarter and perform better than humans in most of the mathematically deterministic fields such as mathematics, physics and some sub fields of biology but it's probabilistically impossible that a text generator with no abstract thinking dimension or systematic intuitive introspections would be better than human cognition fundamentally. I think people referring to AGI actually refer to a 100T parameter model, a gigantic AI model that should technically learn the most subtle and unapparent patterns of its pretraining dataset which even the world class human experts haven't discovered yet.
Me gustaría más bien que usará la inteligencia artificial para ayudar a pagar los incendios más rápidos para localizar a personas que se encuentren en peligro para que se acaben las guerras para hacer tecnologías de siembras y cosas que en vez de quitarlas buscar forma de sembrar y de que los cultivos sean más eficaces y no tener tantos aparatos y tecnologías como para disfrutar mientras el mundo se cae casi quiero que vuelva la era de piedra para que se acabe todo esto
32:00
I was listening to it as entertainment till he said the 1 trillion IQ part, that point he lost me bad...
Why
1 trillion IQ in mathematically deterministic fields only, because no matter the number of parameters there's limits to what the model can learn from its pretraining dataset when the knowledges in that dataset are mostly linear, lately synthetic data have been useful for mathematically deterministic domains such as math, when these models are backed by symbolic engines they can generate high quality synthetic data.
@@TheRealUsername cognition math doesnt even add up like that; you cant take 5 people with 100iq and expect them to perform like a 500 iq. also talking psychometrics, as a science today doesnt even work with those concept, what do we have as an absolute celing for todays metrics? 220iq Terrence Tao? trying to referce orders or magnitude with these metrics parameters its just not really something i can imagine.
5- AI, in order to improve its performance and prevent undesirable consequences, must continuously interact with “effective rules and stable principles in the realm of existence”.
@jamshidi_rahim