bigmadshoe

You don't know anyone who bought storage from google for google photos or drive?

In that case it's just a tradeoff between convenience and security, it isn't because it would be too technically difficult to isolate the systems. Whoever installed them could isolate them if there was no reason for them to be connected to the internet. I doubt most grocery store managers understand the precise engineering behind their freezers either, but they still buy them.

Very likely they were sold on these internet-connected systems without being told about the drawbacks, which is sad.

I think the in the book he was referring to B2C products when he said search will be replaced by AI. Their main B2C product is not ads. I don't think anyone I know has bought an ad from google.

It would be more accurate to say their main revenue stream is ads, which are sold through their main products (search, YouTube, etc.)

You're confusing the probability of the events occurring with the probability of choosing one with the coin flip. Google didn't flip a coin behind the scenes to decide whether to develop AI.

Is it really that hard to isolate some devices from the internet on an existing network?

I would argue that cars are sufficiently advanced now and there are more important things to focus on. I don't care if AI moderately improves the comfort or efficiency of a new car while it took my job.

Obviously we get things in return for the social and economic fallout, carbon emissions, pollution, destruction of education, etc. that AI has provided us. I just haven't heard a compelling argument for why that is nearly enough to warrant this.

The oil is released by the peanuts when you roast and blend them.

Most plain potato chips are like that, not sure how they last but they do.

I found myself confused about what they actually do. They sell you a server rack + software stack that you keep on-prem? The comparison between traditional cloud vs. on-prem vs. oxide is vaguely worded and I can't figure out if their hardware is physically stored on-prem or not.

Maybe I'm just tired!

I'm not sure if you're on a Mac, but I've found the scaling to be a non-issue on my 27" 4K monitor. There are various 'use as' resolution options untied to the actual display resolution (basically: make the UI appear as though I'm running at this lower res, but with higher dpi). For me running 4k with 'use as' set to 1440p is perfect.

What industry are you in? Was the team disbanded specifically because of your productivity with AI?

Interesting, I have the same issue and never thought to change monitor. I think mine is already IPS, but how do you know the quality from the specs? What actually matters for eye fatigue?

> once we're dead, it won't matter anyway

It sounds like something is worth worrying about if you foresee us being dead, presumably prematurely.

LLMs have moved from fancy autocomplete to performing full hours-long tasks in the past two years. The degree to which you are driving the design is being reduced with every model release. At some point they will make almost every design decision, and all that's left is the product stuff in the human world.

I was making two separate points:

A model is not a person -> we need to write new laws. This is not a job for the courts but for us as a society.

The rest of my argument -> information that helps the courts decide, which generally will look at precedent with humans as that is the closest proxy. When you extrapolate from the law as it pertains to humans, the duplication of books for distributed training seems illegal.

Your attitude is extremely sad. Do you think we would be where we are today if oppressed people's throughout history just gave up as easily? We have rights because people fight for them. We collectively have the power to decide what kind of future we want to live in.

Buying and copying one training manual and distributing it to 1000s of human workers is considered illegal, but somehow scanning one book and sending it to 1000s of distributed training instances is not?

Also, you learning something is different than a model learning it, because a model is not a person. You can learn from a book and sell the skills you gained from it, but you can only be in one place at a time. The model can serve that knowledge to every person on the planet simultaneously. We obviously need new laws since this is a fundamentally different situation.

At some point that stops being a "swarm" and just a handful of subagents. E.g. 6 agents running for 30 minutes each isn't really a swarm in my eyes.

But if it was an agent swarm wouldn't we expect something like 300k hours of ChatGPT pro compute equivalent instead of 3?

I don't think you're necessarily wrong today, but AI assisted coding has gone from autocomplete to functioning as a junior dev very quickly. The labs are obviously not happy to leave it at that. Fewer and fewer senior devs will be needed when the model can also create the architecture, derive the tradeoffs, and interpret product requirements directly. I'm really worried about the future of our field at the current pace of things.

Not if the architect said "make a cool design for a house" to some team and went out for the day.

This is really cool! I tried it with my old address (xywz 24th Street) and it doesn't accept it for some reason. I was able to try it from 24th and Mission BART however.

Also "fixed and deployed; your trip should now go Cabrillo → 23rd → Geary". The snappy sentence at the start and the unicode arrows are both pretty big flags for AI writing.

Obviously it's fine to use AI for projects like these, but copy-pasting the response from a coding agent doesn't quite sit right with me.

How could 1. be needed for AI to be general if it is not externally verifiable? I.e. we have no way to know if any system (besides ourself) is conscious.

You're right that we can never be certain a system is generally intelligent by that metric. We can be certain that a system is not generally intelligent, via counter examples.

I imagine it will be like mathematical theorems that have been computationally verified to some very large n. The longer we go without finding a counterexample, the more certain we will be that it's true, even if we can't prove it formally.

The systems today are extremely easy to find counterexamples for, so are obviously not generally intelligent.

The Turing test has nothing to do with consciousness and never has. We still to this day lack a viable consciousness test, as well as an accompanying unifying theory. There have been various attempts to scientifically study consciousness but so far it has eluded us.

I agree with you that we shouldn't limit the scope to economically viable work. But being conscious is not an externally verifiable trait, so there is no way to include a consciousness test when determining if a system is generally intelligent. In fact, you don't know that anyone except yourself is conscious, vs. just pretending. This has already been discussed extensively in the "philosophical zombie" thought experiment: https://en.wikipedia.org/wiki/Philosophical_zombie

There is a general consensus about the shape it would take outside of the grifter community. The general idea is that the system is capable of performing all intellectual tasks that a human is. This does not require being conscious as being conscious is not an externally verifiable attribute, and thus not a task we can measure the performance of a system on.

The broadly used definition says that a system has achieved artificial general intelligence when it matches human performance at most (or more strictly, all) tasks.

Note that being conscious is famously not a task that can be verified by an outside observer. That is to say, even if the task is "describing my conscious experience", a sufficiently intelligent system can simply lie and pretend to be conscious. This concept has already been covered extensively in philosophy separately from AGI discussions, i.e. the concept of a "philosophical zombie" [1]

[1] https://en.wikipedia.org/wiki/Philosophical_zombie

The broadly used definition of AGI has nothing to do with consciousness and consciousness emerging is irrelevant to whether a system can develop AGI.