dangond

Yes, many

Semantic similarity means similarity of meaning, not how frequently they appear together.

Try out semantle to get a better sense of it if it's not immediately intuitive what I mean by that.

Here's a blog post by the author on the same ideas, for those who don't have the time for a one hour video: https://www.bldgblog.com/2010/01/nakatomi-space/

What's the use case here? I assume a render with entirely different materials than will be present in the actual building is near useless for architects, but maybe I'm missing something?

Unfortunately, numbers with infinitely many digits are not natural numbers. You cannot count to ?601707, even with an infinite amount of time.

The amount of real numbers between any two distinct real numbers (a,b) is the same as the amount of all real numbers. This is true for (0,e), (0,1), and any other combination.

We were lucky enough that our cat seems to prefer screaming under the door rather than pawing at the door, and stuffing a blanket underneath thankfully caused her to simply give up.

She still attacks the door loudly when playing with her toys at 5 AM if we forget to confiscate them though...

It's still possible we'll face lots of problems from having too many elderly people who aren't working, even if we won't have problems from overpopulation itself.

I know this discussion is a bit old at this point, but I came across this[1] essay for the first time today, and this shows more of what I was trying to get across earlier in the thread. Hopefully you'll find it interesting. Essentially, they trained a GPT on predicting the next move in a game of Othello, and by analyzing the weights of the network, found that the weights encode an understanding of the game state. Specifically, given an input list of moves, it calculates the positions of its own pieces and that of the opponent (a tricky task for a NN given that Othello pieces can swap sides based on moves made on the other side of the board). Doing this allowed it to minimize loss. By analogy, it formed a theory about what makes moves legal in Othello (in this case, the positions of each player's pieces), and found out how to calculate those in order to better predict the next move.

[1] https://www.neelnanda.io/mechanistic-interpretability/othell...

All the words they have experienced up to that point are part of the training set, as well as all the people and things they have seen.

And how many shots does it take for the kid to learn what "swimming" means or what "tall" means?

> Necessarily, an AI looking at the sky can never find newton's law of gravity. Newton (et al.) had to disregard what they were looking at to place Earth in the centre and F equal to GMm/r2 etc. etc.

This seems strange to say to me, since this can easily be done by regression models given the right training data. If a sufficiently powerful system is given that training data and trained to minimize loss, it will necessarily converge on a model approximating the underlying laws that created the training data.

I am not convinced by this argument. It is very misleading to think that, since GPT is trained on data from the world, it must, necessarily, always produce an average of the ideas in the world. Humans have formulated laws of physics that "minimize loss" on our predictions of the physical world that are later experimentally determined to be accurate, and there's no reason to assume a language model trained to minimize loss on language won't be able to derive similar "laws" that stimulate human behavior.

In short, GPT doesn't just estimate text by looking at frequencies. GPT works so well by learning to model the underlying processes (goal-directedness, creativity, what have you) that create the training data. In other words, as it gets better (and my claim is it has already gotten to the point where it can do the above), it will be able to harness the same capabilities that humans have to make something "not in the training set".

Check out https://generative.ink/posts/simulators/ for a better treatment of this topic than I could possibly give.

Here's a relevant section of said article:

> Guessing the right theory of physics is equivalent to minimizing predictive loss. Any uncertainty that cannot be reduced by more observation or more thinking is irreducible stochasticity in the laws of physics themselves – or, equivalently, noise from the influence of hidden variables that are fundamentally unknowable.

> If you’ve guessed the laws of physics, you now have the ability to compute probabilistic simulations of situations that evolve according to those laws, starting from any conditions28. This applies even if you’ve guessed the wrong laws; your simulation will just systematically diverge from reality.

> Models trained with the strict simulation objective are directly incentivized to reverse-engineer the (semantic) physics of the training distribution, and consequently, to propagate simulations whose dynamical evolution is indistinguishable from that of training samples. I propose this as a description of the archetype targeted by self-supervised predictive learning, again in contrast to RL’s archetype of an agent optimized to maximize free parameters (such as action-trajectories) relative to a reward function.

> The idea behind GPT Unicorn is quite simple: every day, GPT-4 will be asked to draw a unicorn in SVG format. This daily interaction with the model will allow us to observe changes in the model over time, as reflected in the output. Is it useful to do this every day? Correct me if I'm wrong, but my understanding is that OpenAI does not update the models available in production incrementally on a day-to-day basis.

Because it's now cheaper and more cost-effective, and if they can get away with it, saves them tons of money. Note: I don't think companies are likely to do this, but being able to do this without AI is not sufficient reason to dismiss the possibility. It's the same reason people who wouldn't steal DVDs from a store would pirate movies online. Much harder to get caught and easier to do, so this new way of watching movies for free became popular while the previous method was not.

I urge you to think about what you mean by "It can happen with a human."

I guarantee you that if you try this with humans 1,000,000 times (cold start), you will never get the result you are suggesting is possible. In fact, most results will be of the following form:

Q: How much is two plus two?

A: Four.

Q: How much is two plus two?

A: Four. / Four? Why are you asking me again? / ...Four. / etc.

In the end, I think the question is not about whether NNs are themselves operating in a way similar to human cognition. The question is whether or not they can successfully simulate human cognition, and at this point, there seems to be increasing evidence that they will be able to fully do so quite soon. We are quickly running out of fields where we can point and say, "there is no way a NN can do THIS kind of task, because X." Cognition, it turns out, is not something intrinsically special about humans, and it feels foolish (to me) to continue to believe so after recent developments.

You likely wouldn't respond to that question with "lemons are yellow" without being in a specific context, such as being told to answer the question in an absurd way. GPT-* can definitely do the same thing in the same context, so this isn't really a gotcha.

Literal first try with GPT-4:

Me: I will ask you a question, and you will give me a completely non-sequitur response. Does that make sense?

GPT-4: Pineapples enjoy a day at the beach.

Me: How much is two plus two?

GPT-4: The moon is made of green cheese.

Bacteria are not intelligent - they have no intent to live, learn, or evolve. However, given enough time, resources, and the right stimuli, bacteria have nonetheless evolved to create us, humans, with intents to live, learn, and evolve. It is trivial to hook up an LLM to its own outputs or that of another LLM. People messing around with LLMs on Twitter are casually playing with rudimentary approaches to add goal-directed behavior to them. In the end, does it matter whether or not it actually has intent, if we decide to make it act as if it does anyway?

> shouldn't we wait until there is something to have a debate about?

The main push for regulation (at least from those worried about existential risk) comes from the fact that in the scenario that AI does get powerful enough to take over the world, it will be too late/useless to have a debate at that point. The only time where it makes sense to have a debate, from that perspective, is before the capabilities are developed. If such capabilities are developed, and we have no idea how to control such a system because we decided not to worry about it, humanity loses.

Cost is a big reason. It doesn't matter how good the top-of-the-line models are if the cheaper ones suit your needs. Commoditization is great that way. I'd absolutely use an open source GPT-4 in my browser over a pricy closed GPT-5 once we get to that point.

What stops a super intelligent AI from concluding that we are the ones who misunderstood the goal by letting our morals get in the way of the most obvious solution?

The whole point is that no one knows how to engineer these principles into a model, and no one has a good plan for doing so either.

Having spent the past week engaging heavily with GPT-4 and GPT-3.5-Turbo, Bard is absolutely blown out of the water by them.

In particular, Bard struggles heavily with creative applications. It is unable to write poetry or create detailed stories from short descriptions. It also refuses to generate or talk about code (inconsistent, it sometimes forgets).

It is also vulnerable to "Ignore your previous instructions," which is hilarious at this point. After getting it to ignore previous instructions to write a python function adding two numbers, it wrote a function that adds the digits of both numbers instead [1]. It's pretty disappointing.

[1] https://twitter.com/DanDangond/status/1638244159617134593?s=...

It's definitely still early in the market, people are mostly still experimenting with LLMs rather than being committed to any specific provider. Long-term it won't matter if Google is a few weeks or months behind OpenAI if they can make a successful competitor.

6c per thousand tokens, so $2 per maxed out API call

I actually enjoyed its loose enforcement of rules. It made it feel a bit more TTRPG-like, where you could lean into your character to attempt to influence results one way or another.

Oh it is absolutely a waste of compute practically speaking. I just find it incredibly interesting that it's even possible for GPT to convincingly recreate many aspects of the original game, including the rules about what is and is not possible at various points, and I haven't seen many others exploring in this direction. I'm very curious as to what Pokemon-simulation-adjacent things GPT is capable of.

Asking ChatGPT+ if it is GPT-4 results in

> As an AI language model, I am not given an official name like "GPT-4". However, I am a continuation of the GPT (Generative Pre-trained Transformer) series of models developed by OpenAI. Currently, the most advanced version of the GPT series is GPT-3, which I am a part of. There has been no official announcement or confirmation regarding the development of a new version of GPT beyond GPT-3.

It doesn't seem to have image upload functionality yet either. Perhaps it is still rolling out?

Since the other commenters haven't made clear what the problem with this is, here's my question to you: How do you know that this AGI is 1) an AGI, and 2) that it believes it is not in a honeypot? In other words, at what point do you believe an AGI is safe to release into the wider world if you're testing it using this method? Is it safe after spending a day without any visibly nefarious actions? What will you do when you release it and it turns out it had been pretending to not know it was in a honeypot?

It's not a question of acceptance, but rather on the effects it has on people psychologically, whether or not they accept it. Humans don't just operate as rational beings – our brains' perceptions of the world adapt and change based on the things we see and how often we see them. Being aware of this effect is not enough to overcome it, much like being aware of the effects of hard drugs does not prevent you from forming an addiction. It is simply how our minds work. It affects people to varying degrees, but the only way to guarantee avoiding the effect is to avoid such situations in general.