Will AI take your job?

The advisory and consulting and knowledge worker based industries will certainly see an impact - there’s a need for those working there to greatly familiarise themselves with the AI tools available and reassess how they provide client advice and how to continue to bring client value.
As you say, the 80% is there already - what’s the 20% of value add?

One thing I’ve seen is that college graduates on grad programmes are suffering more - no or little promotions - some companies happy to see them leave after a few years- this is something I haven’t seen before and it’s worrying - deep industry knowledge will help existing employees in these sectors - but only for a while .

I think headcount reduction and “doing more with less” will be the first phase of AI impact for some jobs - learning how to use the AI tools effectively to do time consuming tasks, leaving you time to figure out, just what value are you now bringing? 😀

That was part of a marketing campaign which, depending on your point of view about marketing tactics and generating sales leads from rage bait, has been incredibly effective:

https://www.artisan.co/blog/stop-hiring-humans

I hadn’t heard of the term ‘AI boosters’ before, but it’s an apt description which describes people who are invested in ensuring that everyone is talking about AI and imagining it’s the future of the labour market. It’s not even close; it has potential to become a labour saving tool like many innovations in technology have done in the past, but for now as ACD points out, it’s simply more useful for creating disinformation, memes, crap ai art and I’ll add to that, AI slop - AI generated video podcasts are the latest innovation, not an innovation created by AI itself (which would indicate actual intelligence), but rather content which still relies on and requires a user telling it what to do and what is the expected result.

The person behind promoting ‘vibe coding’ as he calls it, also has a vested interest in doing so, but outside of a hobbyist culture, it just doesn’t have any commercial application, other than to generate industry buzz and get Silicon Valley investors hyped about where the next big thing might come from, like crypto, NFTs, etc that can have even bigger fools part with actual money while feeling good about doing so because they’re buying into a concept, which they find attractive:

The concept of vibe coding elaborates on Karpathy's claim from 2023 that "the hottest new programming language is English", meaning that the capabilities of LLMs were such that humans would no longer need to learn specific programming languages to command computers.

A key part of the definition of vibe coding is that the user accepts code without full understanding. Programmer Simon Willison said: "If an LLM wrote every line of your code, but you've reviewed, tested, and understood it all, that's not vibe coding in my book—that's using an LLM as a typing assistant."

https://en.m.wikipedia.org/wiki/Vibe_coding

AI is great at getting “something” down on paper- a very rough first draft - an outline - some suggestions - it will certainly speed up brainstorming of new ideas - but it still needs the human element there to organise it and validate it - but less humans- that’s a given.

As I said, it's welcome to do it.

Imagine when specific AI agents are assigned to create code based on input from huge specialised databases.

This is already the case. ChatGPT is hitting mainstream this year, but ChatGPT is just the tip of the spear. Anywhere that's embracing LLMs has been running agents within large workflows on finetuned models, integrated across every tool you can think of, for quite some time now. The tooling for this is essentially free, extremely powerful, and user friendly. These workflows are completely displacing roles already.

LLMs are also just one small component of the progress. When you look across new modelling approaches, new manufacturing techniques, new use-cases and new integrations, the developments will be exponential now.

E.g. your point on robots, people see Helix independently learning tasks and performing them for an hour, but still seem to think that the state of the art in AI is GPT2.

Anyone thinking that AI will always need human supervision to succeed, in a way that matches current organisation structures, is missing the mark in my view. We're already far past that when you compare the speed and correctness of output to the supervision required. Media commentary and legislation on this is way behind the curve.

If it gets to that point, we'll all be too busy to get jobs. We'll be hiding in rat runs we've dug out that stretch between ruined buildings, trying desperately to avoid being spotted by the killer drones that the billionaire class have built to cull us all, having decided that the general population is no longer of use and must be whittled down to around half a billion.

Well how long has spent with the roadster or semi?

Knock yourself out