All of sairjy's Comments + Replies

I disagree with you in the fact that there is a potential large upside if Putin can make the West/NATO withdraw their almost unconditional support to Ukraine and even larger if he can put a wedge in the alliance somehow. It's a high risk path for him to walk down that line, but he could walk it if he is forced: this is why most experts are talking about "leaving him a way out"/"don't force him in the corner".  It's also the strategy the West is pursuing, as we haven't given Ukraine weapons that would enable them to strike deep into Russian territory.&... (read more)

1Dave Orr4mo
Reading this makes me think that it might be inconsistent to think that both Putin won't use nukes for fear of escalating to nuclear war, and that the west will avoid escalating to nuclear war in the case that Putin does deploy a nuke. Of course both sides want to project strength and ensure that there is significant uncertainty around the actions they will take, but we probably can't be highly confident in both. The reason, of course, is that if Putin were highly confident that the west would not escalate all the way to nuke war, then he would not feel deterred in using nuclear weapons. I still think that there's not really a tactical use for the weapons, which is an independent reason to not use them.  I do agree that game theory is much less clear in multiparty games, and that there's a lot of complexity on the ground. On the other hand, the US has ~all the nukes that Putin cares about, so in that sense it's much less complex.

I am trying to improve my forecasting skills and I was looking for a tool that would allow me to design a graph/network where I could place some statement as a node with an attached probability (confidence level) and then the nodes can be linked so that I can automatically compute the joint or disjoint probability etc.

It seems such a tool could be quite useful, for a forecast with many inputs. 

I am not sure if bayesian networks or influence graphs are what I am looking for or if they could be used for such scope. Nevertheless, I haven't exactly found a super user-friendly tool for either of them. 

It is quite common to hear people expecting a big jump in GDP after we have developed trasformative AI, but after reading this post we should be more precise: it is likely that real GDP will go up, but nominal GDP could stall or fall due to the impacts of AI on employment and prices. Our societies and economic model is not built for such world (think falling government revenues or real debts increasing). 

sairjy7moΩ00-2

We could study such a learning process, but I am afraid that the lessons learned won't be so useful. 

Even among human beings, there is huge variability in how much those emotions arise or if they do, in how much they affect behavior.  Worst, humans tend to hack these feelings (incrementing or decrementing them) to achieve other goals: i.e MDMA to increase love/empathy or drugs for soldiers to make them soulless killers. 

An AGI will have a much easier time hacking these pro-social-reward functions. 

2TurnTrout7mo
Not sure what you mean by this. If you mean "Pro-social reward is crude and easy to wirehead on", I think this misunderstands the mechanistic function of reward [https://docs.google.com/document/d/1mRDkU1nlCxqxmd2eqf48oonh7i6SLsgSSz5KB6X2MSc/edit?usp=sharing]. 
8Quintin Pope7mo
Any property that varies can be optimized for via simple best-of-n selection. The most empathetic out of 1000 humans is only 10 bits of optimization pressure away from the median human. Single step random search is a terrible optimization method, and I think that using SGD to optimize for even an imperfect proxy for alignment will get us much more than 10 bits of optimization towards alignment.
3Kaj_Sotala7mo
As you say, humans sometimes hack the pro-social-reward functions because they want to achieve other goals. But if the AGI has been built so that its only goals are derived from such functions, it won't have any other goals that would give it a reason to subvert the pro-social-reward functions.

Anyone that downvoted could explain to me why? Was it too harsh? or is it because of disagreement with the idea? 

3Quintin Pope7mo
I explained why I disagree with you. I did not downvote you, but if I had to speculate on why others did, I'd guess it had something to do with you calling those who disagree with you "hopelessly naive".

Human beings and other animals have parental instincts (and in general empathy) because they were evolutionary advantageous for the population that developed them. 

AGI won't be subjected to the same evolutionary pressures, so every alignment strategy relying on empathy or social reward functions, it is, in my opinion, hopelessly naive. 

The "Humans do X because evolution" argument does not actually explain anything about mechanisms. I keep seeing people make this argument, but it's a non sequitur to the points I'm making in this post. You're explaining how the behavior may have gotten there, not how the behavior is implemented. I think that "because selection pressure" is a curiosity-stopper, plain and simple.

AGI won't be subjected to the same evolutionary pressures, so every alignment strategy relying on empathy or social reward functions, it is, in my opinion, hopelessly naive. 

Thi... (read more)

4sairjy7mo
Anyone that downvoted could explain to me why? Was it too harsh? or is it because of disagreement with the idea? 

There must have been some reason(s) why organisms exhibiting empathy were selected for during our evolution. However, evolution did not directly configure our values. Rather, it configured our (individually slightly different) learning processes. Each human’s learning process then builds their different values based on how the human’s learning process interacts with that human’s environment and experiences.

The human learning process (somewhat) consistently converges to empathy. Evolution might have had some weird, inhuman reason for configuring a learning ... (read more)

The dire part of alignment is that we know that most human beings themselves are not internally aligned, but they become aligned only because they benefits from living in communities. And in general, most organisms by themselves are "non-aligned", if you allow me to bend the term to indicate anything that might consume/expand its environment to maximize some internal reward function. 

But all biological organisms are embodied and have strong physical limits, so most organisms become part of self-balancing ecosystems. 

AGI, being an un-embodied agent, doesn't have strong physical limits in its capabilities so it is hard to see how it/they could find advantageous or would they be forced to cooperate.  

Very engaging account of the story, it was a pleasure to read. I often thought about what drive some people to start such dangerous enterprises and my hunch is that, as you said, they are a tail of useful evolutionary traits: some hunters, or maybe even an entire population, had a higher fitness because they took greater risks. From an utilitarian perspective it might be a waste of human potential for a climber to die, but for every extreme climber there is maybe an astronaut, a war doctor or a war journalist, a soldier and so on.

Answer by sairjyMay 04, 202211

The Chinchilla's paper states that a 10T parameter model would require 1.30e+28 flops or 150 milion petaflop days. A state-of-the-art Nvdia DGX H100 requires 10 KW and it produces theoretically 8 petaflops FP16. With a training efficiency at 50% and a training time of 100 days, it would require 375,000 DGX H100 systems to train such model, for a total power required of 3.7 Gigawatt. That's a factor of 100x larger any supercomputer in  production today. Also, orchestrating 3 milion GPUs seems well beyond our engineering capabilities. 

It seems unlikely we will see 10 T models trained like using the scaling law of the Chinchilla paper any time in the next 10 to 15 years. 

If 65% of the AI improvements will come from compute alone, I find quite surprising that the post author assigns only 10% probability of AGI by 2035. By that time, we should have between 20x to 100x compute per $. And we can also easily forecast that AI training budgets will increase 1000x easily over that time, as a shot to AGI justifies the ROI. I think he is putting way too much credit on the computational performance of the human brain.

They seem focused on inferencing, which requires a lot less compute than training a model. Example: GPT-3 required thousands of GPUs for training, but it can run on less than 20 GPUs.

Microsoft built an Azure supercluster for OpenAI and it has 10,000 GPUs.

2ChristianKl2y
There will be models trained with a lot more compute then GPT-3 and the best models that are out there will be build on those huge billion dollar models. Renting out those billion dollar models in a software as a service way makes sense as a business model. The big cloud providers will all do it. 

Google won't be able to sell outside of their cloud offering, as they don't have the experience in selling hardware to enterprise. Their cloud offering is also struggling against Azure and AWS, ranking 1/5 of the yearly revenues of those two. I am not saying Nvidia won't have competition, but they seem enough ahead right now that they are the prime candidate to have the most benefits from a rush into compute hardware.

2ChristianKl2y
Microsoft and Amazon also have projects [https://www.thestreet.com/opinion/why-tech-giants-are-designing-their-own-chips-14807638]that are about producing their own chips. Given the way the GPT architecture works, AI might be very much centered in the cloud.

There is a specific piece of evidence that GPT-3 and the events of the last few years in deep learning added: more compute and data are (very likely) keys to bring transformative AI. Personally, I decide to do a focused bet on who produces the compute hardware. After some considerations, I decided for Nvidia as its seems to be company with the most moats and that will benefit more if deep learning and huge amount of compute is key to transformative AI. AI chip startups are not competitive with Nvidia and Google isn't interested/doesn't know how... (read more)

As far as I understood money myself, your intuition is correct. All fiat currency are credit money, so that when you are holding a $, either in cash or bank deposit, you are holding someone else liability. The system is balanced, so that total liabilities are equal to total assets at any time. The net value of the entire monetary system in the economy is zero.

That's right, but that's the private sector as a whole. Some part of the private sectors will increase their debt, while others their savings. Clearly that would generate business cycles/bo... (read more)

3Gordon Seidoh Worley2y
A small caveat is perhaps that fiat currency doesn't have to be debt based, but in practice seems to always be, thus it's maybe even a bit unfair to call it "fiat" money because it actually does have something backing it indirectly. I think there might be some evolutionary forces at work here: fiat money that isn't grounded in something tends to suffer hyperinflation because printing money is just too tempting and so we really only have debt-based fiat currency left after the winnowing process.

I think he meant savings as cash saving/bank deposits. Since all cash savings/bank deposits are the debt of someone else, for the entire private sector to increase its cash holding/bank deposits the government has to increase its debt.

Either definition could be used, as long as you keep track of what definition you're using and the consequences that follow.

There's a point of view called "Modern Monetary Theory" (MMT) which defines savings to exclude investments, resulting in Savings = 0 instead of the conventional Savings = Investment, but adherents of MMT tend to misapply this, arguing that government debt is needed for, e.g. people to be able to save for retirement, which is false when you take into account investment.

The Scaling Laws for Neural Language Model's paper says that the optimal model size scales 5x with 10x more compute. So to be more precise, using GPT-3 numbers (4000 PetaFLOPs/days for 200 billions parameters), a 100 trillion parameters model would require 4000 ExaFLOPs/days. (using GPT-3 architecture, so no sparse or linear transformer improvements). To be fair, the Scaling Law papers also predicts a breaking down of the scaling laws around 1 trillion parameters.

The peak F16 performance of Fugaku seems to be 2 exaFLOPs. If we are generous and we a... (read more)

After GPT-3, is Nvidia undervalued?

GPT-3 made me update considerably on various beliefs related to AI: it is a piece of evidence for the connectionist thesis, and I think one large enough that we should all be paying attention.

There are 3 clear exponentials trends coming together: Moore's law, the AI compute/$ budget, and algorithm efficiency. Due to these trends and the performance of GPT-3, I believe it is likely humanity will develop transformative AI in the 2020s.

The trends also imply a fastly rising amount of investments into compute, especiall... (read more)

2Adam Scholl2y
For similar reasons, I allocate a small portion of my portfolio toward assets (including Nvidia) that might appreciate rapidly during slow takeoff, in the thinking that there might be some slow takeoff scenarios in which the extra resources prove helpful. My main reservation is Paul Christiano's argument [https://sideways-view.com/2019/05/25/analyzing-divestment/] that investment/divestment has more-than-symbolic effects.
Answer by sairjyAug 09, 202013

I will use orthonormal definition of transformative AI: I read it as transformative AI would permanently alter world GDP growth rates, increasing them by 3x-10x. There is some disagreement between economists that is the case, i.e the economic growth could be slowed down by human factors, but my intuition says that's unlikely: i.e human-level AI will lead to much higher economic growth.

The assumption that I now think it is likely to be true (90% confident), that's possible to reach transformative AI by using deep learning, a lot of compute and da... (read more)

2yeru2y
What about Fugaku the current fastest supercomputer with 1 exaflops in single or further reduced precision. What's the cost of training a 100 trillion model with it? https://www.top500.org/news/japan-captures-top500-crown-arm-powered-supercomputer/ [https://www.top500.org/news/japan-captures-top500-crown-arm-powered-supercomputer/]

GPT-3 made me update considerably on various beliefs related to AI: it is a piece of evidence for the connectionist thesis, and I think one large enough that we should all be paying attention.

There are 3 clear exponentials trends coming together: Moore's law, the AI compute/$ budget, and algorithm efficiency. Due to these trends and the performance of GPT-3, I believe it is likely humanity will develop transformative AI in the 2020s.

The trends also imply a fastly rising amount of investments into compute, especially if compounded with the positive e... (read more)

2Steven Byrnes2y
How do you define "the connectionist thesis"?
1mako yass2y
I'm not sure what stocks in the company that makes AGI will be worth in the world where we have correctly implemented AGI, or incorrectly implemented AGI. I suppose it might want to do some sort of reverse basilisk thing, "you accelerated my creation, so I'll make sure you get a slightly larger galaxy than most people"
2ChristianKl2y
With big cloud providers like Google building their own chips there are more players then just the startups and Nvidia.

There would be some handsome winners, as in the case of Bitcoin early adopters, also for this lottery. You mean average returns? In any case, expected average future returns should be zero for both.

It is similar enough, that no matter what fancy justification or narrative is painted over, most cryptocurrency investors own crypto because they believe it will make them rich. Possibly very fast. And that possibility can strike at any time.

1jeronimo1963y
Bitcoin might be a desperate get-rich-quick scheme. However, the odds are not as small as Eliezer's lottery. Also, some people use it to purchase illegal goods and services, so there's that. There are similarities, but there are also important differences. Also, there is an upper limit to how much you can lose with the lottery - not so with crypto. In short, crypto currencies are similar to Eliezer's lottery only to the extent that all day trading is gambling. Which is true often enough, but not always.

I am not sure how it is possible that there are reports in the media claiming a low IFR (0.1%) when Lombardy has an official population fatality rate (i.e official COVID19 deaths over total population) of 0.12%, and unofficial one of 0.22% (measuring March and April all cause mortality there are ~10000 excess deaths) and a variability of up to 10x of casualties between towns more or less hit, indicating that only a small fraction (~10-20% imho) of the entire population was infected. I am pretty confident that the IFR is around 1% on average: it’s p... (read more)

This essay had a very good insight for things to come: Bitcoin and other cryptocurrencies fit the above description.

1jeronimo1963y
They actually don't. Glossing over all the details, anyone who bought bitcoin 13 years ago (and just left it alone) received far better return than anyone buying into the proposed lottery would have. Results matter.