Cool name.
1,557 karma · joined May 8, 2012
https://news.ycombinator.com/item?id=45583970
Thought I might just skip the repetition. You can continue the conversation within that thread.
He found you! I mean, he reached you, and you bought it.
Perhaps that hasn't translated into monetary value, but have you considered the possibility that this was not his goal?
I mean, untapped potential hides some pretty powerful implications. Nature is particularly good at reserving and shaping potential in all sorts of ways that keep it untapped for a lot of time.
The devil is always in the details. For example, in enterprise teams the time to onboard a new project or a new programmer is often dismissed but actually very important. Serious companies take this matter very seriously. More inexperienced teams often come up with an excuse for a slow onboarding time without realizing it's a proxy metric for quality. If you can't jump in quickly or jump someone else quickly into the productive train, something is wrong.
I think recording yourself doing it might be a worthwhile endeavour, even if you don't release it to a wider audience.
The act of recording yourself and watching it might reveal blind spots that you were not aware while doing it. Perhaps things you think are fast but actually took more time than you imagined.
Also, by giving this advice I can take my personal judgement out of the equation. Now it's you recording and only you watching yourself and judging yourself (which can be very soul crushing, I can tell from experience).
There are efforts at estabilishing it though. And it's hard and expensive for wet labs, but it could be much simpler for things like simulating biological pathways.
One could also see your response as "other nations are developing threats, we should race", which I personally think is misguided.
Instead of these petty armchair discussions, we should instead focus on being more serious about it.
https://disarmament.unoda.org/en/our-work/weapons-mass-destr...
You can read more about the efforts for disarmament there.
It goes like this: we assume we're typical life, and we're getting bored of searching, therefore other typical life must also get bored.
It's a challenge to make you think about the point in which such comparisons (life outside must be similar to life here) are true or relevant. It's almost a joke (in the good sense).
This challenge can be expressed by taking the principle that we're not special to the max: if we take it to extend to all aspects of Earth's life, then when we find aliens, they'll be identical to us living in a planet identical to us, which sounds unreasonable.
Therefore, the thing to discuss here is: how special we are and in which ways? How much can we infer about the universe from our current vantage point?
> psychometric
> concrete measurements
So, if you can drive the performance of global education down to a point that the average well-educated adult is as dumb as a brick, you can claim AGI and also cite a paper to support you.
Congrats, folks. You did it.
Also, LLMs have been around for a while. Maybe you can just search for someone that did it and share one video that you would endorse as representative of what you believe to be good context engineering. It seems that there should be a lot of those around, lots of people are using this tech, aren't they?
The only thing to be said about it that resonates with what I'm concerned with is that anyone that is good in the head wants better international oversight on potential bioweapons development.
I do not endorse the view that covid was engineered. Also, I consider it to be unrelated to what I am concerned about, and I will kindly explain it to you:
Traditional labs work with the wet stuff. And there are a lot of safeguards (the levels you mentioned didn't came out of thin air). Of course I am in favor of enforcing the existing safeguards to the most ethical levels possible.
However, when I say that I am concerned about AI being used to circumvent international agreements, I am talking about loopholes that could allow progress in the development of bioweapons without the use of wet labs. For example, by carefully weaving around international rules and doing the development using simulations, which can bypass outdated assumptions that didn't foresaw that this could be possible when they were conceived.
This is not new. For example, many people were concerned about research on fusion energy related to compressing fuel pellets, which could be seen as a way of weaving around international treatises on the development of precursor components to more powerful nuclear weapons (better triggers, smaller warheads, all kinds of nasty things).
For example, it could be that some company offered this kind of AI to the army, with those exact words, and they refused (which sounds incredibly reasonable). The interview would serve a purpose of deliberately mocking such offer, in public.
There is almost no downside to this, as you already mentioned "everybody and their brother predicts Military Involvement [with AI]", and I quote you to illustrate that the idea that this is engrained in the popular culture (maybe by design).
Companies such as Raytheon and other traditional military contractors have been researching AI (not necessarily chats) way before Sillicon Valley. They can (and probably do) offer consultancy to the army regarding which technologies are safe to employ for critical decision-making.
I'm not defending them, by the way, just exploring some ideas around what possibly is going on here. Letting my imagination go a little beyond what the typical scenarios imply.
What I know for sure, is that AI-2027 is not a serious prediction. It's a piece of fiction. Kind of in the vein of The Blair Witch, with some tones of realism to impress the public, and a little bit more sinister in the sense that it resists the idea of being interpreted as fiction (which I think it's a downer, they could have reached more acclaim by being artists than scammers).
Why do people believe in the AI-2027 timeline and not Terminator 3: Rise of The Machines?
That's because AI-2027 is not labeled as fiction, the categorization it actually deserves.
Fiction can sometimes have predictive power, but that's not the case for AI-2027. Their goal does not seem to be creating art to entertain people.
Technical analysis is not good enough. It hides some tricks these companies seem to be relying upon.
If you buy/manufacture a bleeding edge server chip, and you don't use it, then you can resell it or repurpose it into an end consumer product.
I already explained this. It is what happened with Xeon processors. Lots of obsolete server chips that chinese manufacturers found a way to repurpose into competitive mid-tier end user machines. I think that's a significant event that ties in directly to the bubble we're seeing.
I also already explained why I believe M4 mac minis are a product of a similar practice. A little less obscene because it's mostly inside the same company, with no way of selling the same chip twice to different buyers.
These new deals are just the next iteration of this trick, in which the surpluses of overscaling are being dealt beforehand instead of waiting for them to become e-waste.
I also believe the confidence these companies have in this trick is overestimated. It works for Apple (for now) because they have a loyal customer base willing to buy newer versions even if the new versions are not that better than the previous.
This also explains why the cycles of hardware generations are getting shorter.
The application layer accounts for what percentage of produced components (the max they can build, presumably) will be used (instead of saved for being repurposed). It's not holding the whole thing together.
The more I look into it, the more it seems that it's going that way. For example, the Broadcom deal could be a sign of guaranteeing a supply of wifi chips, necessary to make end user products from the obsolete server parts that were not used in datacenters.
Again, I could be wrong, but I don't think I am. The predictions I'm making for these moves can be tested with time, and we will see what happens. There are ways it can go terribly wrong.
I decided to narrow the discussion to focus on being accurate, as you demanded. It surprises me that now you're not interested in it.
Humans get rewarded for thinking "I don't know", a lot. That's why it's hard to compare.
> A model that always bluffs
A model doesn't bluff. It feels to us humans that they bluff, but there is no bluff mechanics in play. The model doesn't assess the prompter's ability to call their bluff. It's not hiding that it doesn't know something. It's just not reached a predictable point in a sequence of token predictions that can or not have something that resembles a call to what resembles a bluff.
Up to the point it's corrected, the model's representation of what was asked is the best it can do. It has no means to judge itself. Which leads to...
> The real issue isn’t that models make things up; it’s that they don’t clearly signal how confident they are when they do.
Which sounds like exactly what I said, but it's not. Signaling confidence is just a more convincing faux-bluff. Signaling is a side-effect of bluffing, a symptom, not the real thing (which is more related to asessing whoever is on the other side of the conversation).
> Imagining things, seeing problems from the wrong angle, and even fabricating explanations are the seeds of creativity.
I agree with this. However, Newton was not bluffing, he was right and confident about it, and right about being confident about it. It just turns out that his description was of a lesser knowledge resolution than Einsten's.
For this to work, we need lots of "connective tissue" ideas. Roads we can explore freely without being called liars. Things we can say without saying that these things are true or false, without the need for being confident or right, without being assessed directly. This is outside the realm of bluffing or saying useful things. It's quite the opposite.
When people saw comets and described them as dragons in the sky, they were not hallucinating or telling lies, they were preserving some connective tissue idea the best they could, outside of the realm of being right or wrong. This were not bluffs. There were some "truths" about their mistakes, or something useful (they were unadvertedly recording astronomical data, before astronomy existed). Those humans felt that was important, those stories stuck. Can we say the same thing about LLM hallucinations? I don't think we're ready to answer that.
So, yes. Hallucinations could be a feature, but there's a lot missing here.
> People do project their general behaviour patterns on social media regardless of the form
I need sources (not examples!) that demonstrate that people project their general behavioral patterns on social media.
I also need sources (not examples!) that demonstrate that this projection remains the same regardless of the form (video, text, etc).
Your line of reasoning implies that Twitter was one of the means for people to do that, and the reach lost by the dissolution of Twitter communities exemplifies this projection mechanism. It's a fair assumption, but not enough to prove that *general* behavior was present in such interations, nor that it works *regardless of form*. So, there's a lot to prove.
If I were to guess, I would say the case is much more specific, and the generalization does not hold water.
Good luck.
It sounds anti-LLM, but it actually helps support the illusion that LLMs can do more than they actually can.
I don't think an LLM can write serious software on its own. If it could, there would be some extraordinary evidence, but all there is are some people spreading rumours. If you ask them for simple evidence of comparable performance (like a video), they shy away or answer vaguely.
The thing is not there yet, and I understand the optimism of some, but I also must emphasize that it's not looking great for LLM coding enthusiasts right now. There's no amount of proselitism that can make up for the lack of substance in their claims. Maybe they can trick investors and some kids, but that's not going to cut it in the long run.
Therefore, this is not a problem. I don't need to worry about it. If (or when) some evidence appears, I can then worry about it. This hasn't happened yet.
Trump somehow contributed to that radicalization (he is not the sole responsible for it though). I think many who support him don't see it that way, and instead interpret his actions as some sort of cheat codes for progress that cut through a lot of bullshit. I am skeptic of his agenda.
I am, however, also cautious about many of the leftist ideas, specially in the last decade. If made to chose though, I would definitely go left.
I think I'm an orphan of a deceased left that doesn't exist anymore. A left that cares more about things like education and healthcare than about how they look on social media. I don't see much value in discussing leaders (this comment is a rare exception) or amplifying partisan narratives.
For example, this post could have been a product of just probing a particular group of people to understand if they are interested in the subject and what they have to say about it.
That can be done indirectly, by suggesting someone (offering a link or planting an idea) that is already known to be interested in surveillance and prone to share interesting discoveries (in other words, the poster might not even be aware he could be an asset).
Think about the many ways someone could know your interests and how prone you are to react to something and how that could be used. If you are in tech, think about all the silly ways that kind of information can leak publicly.
People often disregard the possibility that they could be an active part of a surveillance network (as an unkowingly asset), instead focusing on more fantastical ideas such as technological hacks or coding wizardry.