Artificial Intelligence and the Future of Work

I think this is correct and also the point.

Neural networks, deep learning modes, have been reliably improving year to year for a very long time. Even in the 90's on CPUs, the combination of CPU improvements and training algorithm improvements translated into a noticeable upward arc.

However, they were not yet suitable for anything but small boutique problems. The computing power, speed, RAM, etc. just wasn't there until GPU computing took off.

Since then, compounding GPU power, and relatively simple changes in architecture have let deep learning rapidly become relevant in ... well every field where data is relevant. And progress has not just been reliable, but noticeably accelerated every few years over the last two decades.

So while you are right, today's AI varies from interesting, to moderately helpful but not Earth shattering in many areas, that is what happens when a new wave of technology crosses the threshold of usability.

Past example: "Cars are really not much better than horses, and very finicky." But the cars were on a completely different arc of improvement.

The limitations of current AI models aside, their generality of expertise (flawed as it might be), is unprecedented. Multi-modal systems, longer context windows, and systems for improving glitchy behavior are a given, and will make big quality differences. Those are obvious requirements with relatively obvious means.

We are going to get more than that going forward, just as these models have often been surprisingly useful (at much lower levels and narrower contexts) in the far and recent past.

This train has been accelerating for over three and a half decades. It isn't going to suddenly slow down because it just passed "Go". The opposite.

I believe that there is no "one size fits all" answer to this, and failing to get in front of that in discussion, harms the resulting thought. The societal situation of an individual using their talent to make arts or design, alone or in self-selected teams, is not the same as large companies who run market systems and have attorneys and accountants to aid that over time.

Lots of excited people argue against copyright and then go directly to the story of the Mickey Mouse image.

It is precisely because copying is so, so different than creating, that the situations within the breadth of this topic are not, and should not, be comparable.

"A great society is judged by how it treats the least of its members" very much applies to working arts and crafts adults IMHO.

I think you're selling "it is about jobs, not data" a little short here.

Let me start off by saying I work on AI for healthcare, my first IRB protocol that contemplated computer vision is from 2012. I'm not an OG, but I've been working on this stuff for a while, and I'm very bullish.

On the flipside, my wife is a pediatric occupational therapist, she works with autistic kids. Kids in the Bay Area. Her clients are Google machine learning engineers. The major issue with jobs is not so much the monetary value of the work, although that's an important secondary outcome.

Humans need an occupation. Occupation is the purpose of life. Long before money was a thing, we needed purpose. Even kings and their courtiers, before money, needed occupations. We need to be doing something. People change their occupations from time to time. It doesn't have to be a job. It could be a hobby, in some cases. Even kids want to contribute. Even infants, as soon as they understand and can, will reach out to console or delight their caregiver. But we need to feel like we are giving back. Our minds and muscles atrophy if we are not occupied.

If we attach everything to solar power and let the machines run the world, we'll end up some version of the blob people in Wall-E's Buy n Large spaceships. The reward at the end of the movie was the people getting off the space ship, thanking the robots for giving them back purpose, mainly to restore their planet.

Purpose and occupation are about so much more than money.

Using the unmonitored output of a LLM-translation service for your commercial content, outputting in languages you can't read, represents a big reduction in quality assurance and greatly increases the risk of brand embarrassment and possibly even product misrepresentation, while leaving you with no recourse to blame or shift liability.

> If I was a vendor whose business was at all centered around language or data ETL or anything that involves taking text and doing something with it, I would be absolutely terrified at someone writing a 20-line python script with a good system prompt that would make my entire business's reason for being evaporate.

The more likely future is that existing translation houses will increasingly turn to LLM-assistance to raise the efficiency and lower the skill threshold for their staff, who still deliver the actual key values of quality assurance and accountability. This will likely drive prices down and greatly reduce how many people are working as translators in these firms, but it's an opportunity for them, not a threat.

LLM's don't seem to be on track to be the foolproof end-user tools that the earyl hype promised. They don't let us magically do everything ourselves and (like crypto being imcompatible with necessary regulations), they don't offer all the other assurances that orgs need when they hire vendors. But they can very likely accelerate trained people in certain cases and still have an impact on industry through specialty vendors that build internal workflows around them.

Yeah, no.

As part of a hilariously bad set of actions by a corporation that I had to threaten with legal action, I decided to try seeing what ChatGPT had to say, knowing in advance all the problems with it in this field, and… it was pretty much what I expected — enough to be interesting and get the general vibe right, but basically every specific detail that I could look up independently without legal training of my own, were citations of things that didn't exist.

I'd just about trust them on language tutoring, but even then only on the best supported languages.

Use them as enthusiastic free interns-to-juniors depending on the field. At some point, they'll be better, but not in predictable ways or on a predictable schedule.

But they are pretty general in their abilities — not perfectly general, but pretty general — so when they can do any of what you've suggested to a standard where they can make those categories unemployable, they're likely to perform all other text-based (not language-based, has to be text but doesn't have to be words) economic activity to a similar standard.