Dario Amodei, CEO of AI firm Anthropic, in October 2024 penned an optimistic vision of the future when AI and robots can do most work in a 14,000 word essay entitled – 'Machines of Loving Grace'.
Last month Mr Amodei was reported as saying the following – “I don’t know exactly when it’ll come,” CEO Dario Amodei told the Wall Street Journal. “I don’t know if it’ll be 2027…I don’t think it will be a whole bunch longer than that when AI systems are better than humans at almost everything. Better than almost all humans at almost everything. And then eventually better than all humans at everything.”
Although Mr Amodei wasn't present at the recent inauguration, the rest of Big Tech was. They seem united behind America's most prominent South African, in his bid to tear down the American administrative state and remake it (into who knows what?). Simultaneously they are leading us into a future where we will have to compete with robots & AI for jobs, where they are better than us, and cost pennies an hour to employ.
Mr. Amodei is rapidly making this world of non-human workers come true, but at least he has a vision for what comes after. What about the rest of Big Tech? How long can they just preach the virtues of destruction, but not tell us what will arise from the ashes afterwards?
AI belonging to Anthropic, who's CEO penned the optimistic 'Machines of Loving Grace', just automated away 40% of software engineering work on a leading freelancer platform.
byu/lughnasadh inFuturology
9 Comments
Sure, the economies will tank and we will become their slaves without anything besides menial work to perform. Or? What’s your solution?
They didn’t actually do this.
The paper seems to indicate that they scraped the job requests and had an AI propose solutions, including for jobs that were listed for $50. They had software engineers write end-to-end tests for a solution and then compared the LLM’s solution to the E2E tests and found that it could have solved many of them.
We know LLMs can solve a lot of coding issues or present solutions for existing problems, especially if the problems are “easily testable” (which they admit is a bias in their data).
I’m not saying the day isn’t coming where LLMs can literally just take tasks from Upwork and do them (which would effectively cut out upwork since you would only need the AI), but in this instance it was a speculative test with a lot of biases; the LLM didn’t actually earn any money.
Man, Machines Of Loving Grace was one of my favorite bands of the ’90s. I’m gonna go hop in a time machine to a decade far removed from the bizarre future we keep finding ourselves rapidly sliding toward.
> For high-value IC SWE tasks (with payout exceeding $5,000), a team of ten experienced engineers validated each task, confirming that the environment was properly configured and test coverage was robust.
You too can automate low-level tasks with the help of 10 experienced engineers making sure that the task is easily automatable and then writing significant numbers of frontend tests!
Folks who have done this sort of freelancing before know that a lot of the tasks – especially for open source software like Expensify, tend to be the kind of things you’d give an “integration engineer”. They tend to be extremely finite and often not novel.
This remains unconvincing as evidence that LLMs can do any level of software engineering.
The paper literally says that models fail most of these challenges, what am I missing?
Laid-off engineers need to be building tools that will dismantle the AI tools.
Clearly the goal is to eliminate labor so a few billionaires can profit.
The title of the text comes from a Richard Brautigan poem, I believe.
Obviously it sucks if a bunch of people get laid off, but this means that products are getting cheaper to make, and when there’s market competition, over time this makes things cheaper. Music is essentially free now, for example, since it’s so cheap to distribute online.
And of course, there will always be jobs designing, building, moving, repairing, and maintaining the machines that do things for us. And if the machines do that, then everything will be free. And if someone still tries to charge money when people don’t have jobs, people will make and trade things with each other, remaking our current non-AI economy.
BTW “America’s most prominent South African” is nowhere near the forefront of AI and just has an also-ran company, not sure what that has to do with this.