Rational Aussie @rationalaussie

@Travis_Kling 💯

See:

Rational Aussie @rationalaussie

6 days ago

@DanXLancaster If you have to compress the cognition so much so that it's computationally reducible to the limits of our brain, that explanation is extremely lossy in a way that it would not be to a superintelligence. It's like how you would answer a child's question about why the sky is blue.

1 0 5 781 0

The best software engineers are not reviewing code anymore, they are building the system (AI) that reviews the code, I.e. the guardrail system. Do they 'understand' the code at any significant level? No, not really. Could they, if they asked AI to explain it to them? Perhaps, today. Just. But there are a couple of problems with this. Firstly you are relying on the system under testing to tell you whether the system itself is safe. You don't really know whether the AI is aligned, has the right context, or whether you've framed the question right in order for its explanation to assist you find the truth. And secondly, complexity is going to scale as AI scales. Meaning it will take more and more human hours to truly 'understand' what the system does. This is already what's happening in software, and it will happen everywhere else too eventually. You only need to look at the OpenAI Hugging Face incident, and how they 'debugged' it, to see how human understanding will not persist in any true sense.

Daniel Lancaster @DanXLancaster

6 days ago

@rationalaussie That AI's in general will be better managing these fleets, yes, no disagreement. As to jobs being so 'complex that humans do not even understand nor can assess them' - I think this is wrong. There are, per Deutsch and Dennett, no explanations beyond our cognition.

1 0 1 4K 1

If you have to compress the cognition so much so that it's computationally reducible to the limits of our brain, that explanation is extremely lossy in a way that it would not be to a superintelligence. It's like how you would answer a child's question about why the sky is blue. What level of detail are you gonna go into? What prerequisites must they have to truly understand the true explanation? Quite a bit, so you compress it such that they can 'understand' what you are saying. Very lossy.

What makes you think a human will be better able to manage a fleet of superintelligent AIs doing entire jobs so complex that humans do not even understand nor can assess them, versus another superintelligent AI managing them instead? Your logic rests on this, but empirically we are already seeing loss of human control on the default path through sheer incentives. So much digital output is being produced that humans are unable to assess its veracity, so that is increasingly being done by AIs too. This idea that humans will remain in the loop managing AIs over the long term makes no sense. A species of superior power, influence and intelligence does not merely find itself managed by one dumber than it by orders of magnitude. The system as a whole will perform significantly WORSE if the human is in the loop, not better. Anyone who keeps the human will operate slower, dumber, and with materially higher costs. Without regulation requiring it, that's not where the economic incentives lead us - they lead the opposite way.

Pierre Rochard @BitcoinPierre

6 days ago

@gmiller AI is going to make human labor more valuable because human labor and productive capital are complementary. Give a worker control over vastly more capital and you increase the value of that worker’s judgment, decisions, and effort. Make that capital widely available and competing

10 7 51 9K 48

@InfamousLetters If you keep the people and they are both accountable and do not understand, then what? They are blamed for a system that is cognitively impossible to grasp?

Shitposting (the OG UBI)

username @username566494

7 days ago

@rationalaussie What should a knowledge workers reposition into?

0 0 0 1K 0

@Georgism @Olking07 Good post. Not really sold on LVT though. How would it work? Doesn't it assume the people that own that land are liquid and employed, which is not necessarily going to be the case?

@E_IS_Ferguson There will be guardrails and heavy regulations around them.

Imagine for a moment that it's 2032 and your recently purchased humanoid robot finally arrives at your dogbox. It walks itself right up the stairs to your door, knocks, and walks in. It can cook worldclass meals for you, teach you anything you're curious about, go for a run with you, cut your hair, be your physio if you get injured, and punch you in the face if you spend too much time doomscrolling (then act as your psychologist when you have withdrawl effects). It can hop in a self driving car and build you a cabin in the woods if that's your thing. It's smarter than you at every field ever conceived by a ridiculous margin, millions of times faster than you, and significantly cheaper than you in terms of its task completed per unit of energy. Truly imagine that, then tell me how the modern economy - the one where a human is supposed to work a full-time job in exchange for money to survive - persists. You are dreaming. From first principles it makes no sense. We are rapidly approaching a discontinuity - a Singularity. Existing institutions will not survive. Our way of life will change. Hopefully for the better.

@TMTLongShort Absolutely

It's a real shame we are in the final innings of 'the laptop job', but clicking around on a computer screen really ain't it long term unfortunately. And by long-term I mean 2-3 years max.

@JHochderffer Humans are not special. We will be seen as completely retarded by superintelligence whose only real limit is the laws of physics.

1. The ChatGPT for robotics moment will happen next year. 2. By the 2030 humanoids will begin rolling out to homes. By 2035 they will be in most homes across most developed countries. We will experience a recursive industrial explosion of robots building robots, combined with superintelligence directing resource/capital allocation and coordination causing everything to go way faster than most think. 3. Disagree on timeline, disagree on net new job creation over the timeline that matters, and disagree humans will have any economic utility in this new world. 4. That's cool, but it should be decoupled from someone's ability to survive lest we all want to live in a world where there are 1000 yogo studios competing in one suburb because relational skills are all that's left to monetise. Why would a human be any better at directing, monitoring and enjoying the output than a superintelligence? It won't. There's no economic utility from it. Pleasure and leisure, sure. That's a different system though, and capitalism will not survive this transition.

@BenIsBack123 I tend to agree. Humans have a strong disgust sensitivity to useless people due to the way we evolved, and unfortunately AI makes most people economically useless in the long run.

@btud01 I agree it's all problematic. People will have to decide based on their own personal circumstances what the right position for them is. But blind ignorance from some of what's happening isn't a good response

You're not extrapolating the exponential properly. Human digital workers will become negative marginal value over the next few years. They will be competing with AI digital workers who are much faster, smarter and cheaper than them. It doesn't matter how long it takes to diffuse fully, the trajectory is clear. Yeah it might take a decade to diffuse, but you know what happens after those 10 years? There's virtually no pure play laptop jobs. The entire knowledge economy will be nuked. Winning at life is largely about positioning: 1) did you see what was coming and get it right directionally, and 2) did you size your bet correctly? If the Titanic is sinking do you stay on it until the very end just because it hasn't sunk yet, or do you abandon ship immediately and reposition elsewhere before everyone else realises there's only a limited number of lifeboats? Winners win because they are willing to cut all losers and reposition whilst everyone else's ego prevents them from doing so. And they size their bets correctly - no hedging if they know what is inevitable. 'Muh laptop job will persist until 2034 but I'll be paid 34 cents a week and work 90 hours a week' is the equivalent of what some guys saying 'this will take a long time to play out!' are really not understanding. Every year you don't reposition your position gets worse, and you feel more dug in to something that is structurally broken from here on out.

The Bogan Economist @BoganEconomist

a week ago

@rationalaussie Mate you have great insight but your timeline is wayyyy off! Look at the adoption of big technology shifts over the years. Then you will understand everything you think should happen, will take about 15-20 years before it actually does.

5 0 8 19K 4

@nicola_bah41160 *turns Opus 5.5. to Ultracode mode* "Claude, write me the most optimal digital signal processing algorithm in human history." That's all it takes

@AbHomineDeus 💯