Hunar Ahmad

MD, PhD Neuroscience.
Ambition level: 1000%
Success rate: 0.1%

[Hunar Ahmad @hunar012]

I made this fun side project and discovered many interesting self organizing patterns! The code is less than a page (excluding GUI). #artificial #life #simulation

[Hunar Ahmad @hunar012]

@fchollet I think the word is modularity. Modular software has more boilerplate code at the expense of scalability. Transformers are more modular than naive NN, hence, more scalable. Modularity = Generalizability.

[Hunar Ahmad @hunar012]

@bigclusters @LaurenceRh62785 No, it's not me!

[Hunar Ahmad @hunar012]

@absurdloops Thanks. I'm glad my videos were useful.

[Hunar Ahmad @hunar012]

@victortrac @Andercot Even with a universe-sized computer, you can only predict the next state of the universe or whatever is the smallest time-step update (you can't jump to any point in future because of computational irreduciblity)

[Hunar Ahmad @hunar012]

@goddek @elonmusk What if fixing Earth depends on going to Mars first

[Hunar Ahmad @hunar012]

@Plinz Thanks for your input. This also reminds me of another topic which I think is related - The Localist vs. the Distributed representation of connectionist models. The latter looks to be the winner in today's approach for AI.

[Hunar Ahmad @hunar012]

@geoffreyhinton @ylecun Stuck at a local maximamum

[Hunar Ahmad @hunar012]

@karpathy You don't need to work for any company. Do educational YT videos. You love education and people love you.

[Hunar Ahmad @hunar012]

@elonmusk But why hate did evolve in the first place? It must have some survival value in some situations especially where fear is involved. So beating hate = beating fear

[Hunar Ahmad @hunar012]

@ilyasut Intelligence evolved later in life. So yes there are many other animal qualities have a much higher value for the survival in evolutionary terms.

[Hunar Ahmad @hunar012]

@tegmark But LLM's are trained on Wikipedia and Wikipedia pages have the coordinates of every city. It would be interesting if the model can predict the coordinates of locations not in their training data.

[Hunar Ahmad @hunar012]

@pickover Step 3 must be wrong. Here is a less confusing solution:

[Hunar Ahmad @hunar012]

@NPCollapse But the guy is a smart businessman. Many of his words seems to me as a smart advert for open AI's ability and value.

[Hunar Ahmad @hunar012]

@JoeBathelt @bunq Sorry to hear that and thanks for sharing your experience. It seems to me that your bank's security is grossly negligent.

[Hunar Ahmad @hunar012]

@zaesarius Yes, I remember reading this few years ago. It's an easy read and enjoyable. Nowadays I prefer more technical books🤓

[Hunar Ahmad @hunar012]

@Plinz The feeling of continuity might be an illusion in an infinite possibility universe. Even If the universe change its state randomly every second, we would still feel continuity if we have false past memories in each of the independent random state changes.

[Hunar Ahmad @hunar012]

@Plinz They stuck on a local maximum. Prompting them to leave local maxima is not easy.

[Hunar Ahmad @hunar012]

@lexfridman Only 2 eyes become blind!
However, "Two eyes for an eye makes the whole world blind." -Gandhi with a math degree

[Hunar Ahmad @hunar012]

@elonmusk @xai Hope we truly use AI to understand reality instead of using it for business or to fuel our primitive addictions.