~10 years ago I worked on having a “cab” on my PC learn the most efficient route from pick up to drop off at the airport…as it entered its environment it would run off the road -a penalty…I’ll never forget my excitement when it taught itself the best route…we’ve come a very long way…

Hi Tina, thank you for sharing. Keep up!

Awesome video!!!

"Great video! The Mouse and Maze example provided an excellent demonstration of the Q* Algorithm in action within AI. It’s fascinating to see how, akin to a mouse skillfully navigating a maze, the algorithm effectively traverses through complex problems to identify the optimal solution. I believe the analogy with uranium enrichment is quite apt. Similar to how uranium enrichment concentrates U-235 for nuclear energy or weaponry, the Q* Algorithm zeroes in on the most efficient strategies in AI, significantly improving decision-making processes. This comparison not only highlights the effectiveness of Q* in artificial intelligence but also sheds light on its potential impact and significance. Eagerly anticipating more of such enlightening content! Thank you, Vikas

Awesome content ❤

Thank you

Thank you ❤

Great explanations! Surprised Food Wars reference didnt show up 😉

I really like her Videos, it is super challenging to make video attractive to wide range of audience... for those who who has not much idea about AI and ML and also for people who do

@TinaHuang1

5 ай бұрын

awww ty I try 🥺

Tina, why don't you put up popcorn warnings at the beginning of your videos so that the viers have time to pause and go make some popcorn before watching you super leet content. Soooo goood! Soooo well put together, researched, and explained. Thank you

AI is exciting but scary. It's not just a science fiction gimmick or some distant event that might happen. It's real and happening right now.

Q* sounds like the algo they used many years ago to build a chessplayer that no human could beat. There was no row then about how scary it was. I imagine a customer support that is polite in such a creeping way as to make people super aggressive when they do not get the right answer! It is like when the phrase ”have a nice day” nearly became an insult.

well, one thing to keep in mind is that in the grander scheme of things, all life must end, eventually in heat death, and we humans in particular might be much much earlier than that, through nuclear war or extreme global warming. So, in this context, if an AGI is how we end, I'm pretty pleased. Obviously I still prefer if we live on to become galactic federations where individuals enjoy thousand years lives and extreme comfort, but history tells me that's basically never gonna happen. So, in the choice between nuclear war and climate change, I'd choose a rogue AI to end us all.

@15:44 humans also make mistakes. The more i use and build with AI the more I see the "flaws" as temporary. I think we can build reasoning into these machines with enough rounds of reinforcement learning. Hell...train it on every book/lecture on formal logic and reasoning we have. A fun project could be to take Llama 7B (or a more uncensored open source model) and fine-tune it on copius amounts of formal logic books/lectures. And keep retraining until we get satisfactory reasoning capabilities

@ObsceneSuperMatt

5 ай бұрын

As long as it has the ability to unlearn erroneous information. Otherwise the flaws can permanently taint the outputs.

how is synthetic data does not cause overfitting? how can it have entropy?

As AI's development accelerates beyond our control, I'm not scared a tiny bit about its capabilities or advancements. What I am afraid of is humans who make decisions about the AIs and their integrations. Just listening to them speak, it's become apparent that some at the very high are mistaking models for reality or for even Reality, and synthetic data for the real world data. That scares me.

Im am laying in my bed with four kittens on me and they are all meowing ”AI is a piece of cheese!”

Sam did an interview on Trevor Noah's podcast yesterday that seems to be under most people's radar. It has full audio and video on Spotify. No mention of Ilya, or Q that I noticed, but interesting nonetheless. Kind of low energy, in my opinion.

Serious question - why is it being called a discovery and not an invention?

My conspiracy brain is telling me that Google's discovery and release of 700 new materials indicate that they got to Q* a lot earlier than OpenAI and have already retrained their AI with Q* methodology (fully synthetically, and thus the results they are generating)

69% nice!

🥰

Gemini just came in and absolutely destroyed GPT-4 we n e e d a vid on that 🙏

Whenever I tried to get help solving integrals AI made me pull ny hair out! Videos from humans way better for now

Illya is a good guy. Sam is true evil but all the hatred was give to illya

First!

What are sam altman's academic bona fides? He's making quite a lot of claims. Great claims require Great evidence

Nope

On another episode of AI Doomer:

you cute :p