State of the Market

[SussexSeagull]

I highly doubt it. As said above, accountability.

Someone in energy trading, makes a trade based on AI and it's incorrect, his manager or director asks, why and the response was 'AI told me.' It will not go down well.

You think Defense relies on AI for drones?

AI, artificial intelligence. The clue is in the name. Artificial.

When AI becomes good enough to predict sporting outcomes and gets banned by bookies, I will invest time and effort into it, until then, it's probably better than just using Google.

There are already data firms such as StarLizard (owned by Brighton and Hove Albion Chair Tony Bloom) who try and beat the bookies for large value customer. AI couldn't predict things like a star striker waking up on the morning of a match with a cold (as his club are unlikely to let anyone else know if there is a cold outbreak amongst the squad).

Not really my field but drones are starting to be used as a wingman for a human pilot who hasn't got time to control a drone as well as pilot a plane. The next phase will be multiple drones, so essentially a squadron with only one human being in the air and support back at base.

[SchumiStars]

Anyway, 2025, Feb 2025, coming up. Market is tulip.

Just been cycling and going gym. At least, I don't look like a software dev. That's for all the C++ guys who have beards and walk around in sandals

Although at this stage, I would be quite happy to change my appearance for a contract.

[Snooky]

I think you are right but some are getting a bit carried away thinking they just need to fire up ChatGPT and get rid of their Software Development department.

Agreed. I use ChatGPT & CoPilot regularly to suggest solutions for issues in things like Python, SQL, Unix and Excel. I could work them all out myself but why waste the time doing that?

In quite a few instances, the AI response is just plain wrong, in terms of either logic or actual valid code, or it ignores some of my criteria. When I mention this to the AI bot, it apologises and admits I'm right, then tries again. But I've had cases where it just flips back & forth between two incorrect solutions and I have to go off and spend time working out the correct solution myself. A novice in that technical area might not notice that it had given an invalid answer.

I guess my point is that current publicly available AI solutions can't tell definitively whether they're right or wrong, so "intelligence" is probably the wrong noun for it at present. While that's the case, trying to replace experienced humans with AI is a very risky gamble.

[siddhantkumar]

Genuinely interested in this, maybe a business thread in it's own right.....

is your client using ChatGPT, CoPilot behind the scenes as it were or is it humans\traders typing in prompts (the only I know how to use them)?

Is AI being used as a bit of a catch all term? Before AI I would have thought defence could already identify (image processing\matching) and track (radar) flying things? Haven't FMCG's used data analysis\maths to predict demand\supply before AI?

Like I said genuinely (not trolling) interested.

Artificial intelligence is not new and definitely not post OpenAI. It has been used a lot in energy sector (that I know off) quite a lot to train neural networks by feeding data on weather patterns, sports events and what not, to predict energy requirements over 6 months, a week, day ahead or even in next 15 min.

training on data, looking at old data plus predicting future product demand based on various other information is base for AI use in retail.

edit: developers are using it in visual studio, traders in their excel and python to model demand supply (or whatever they do)

[siddhantkumar]

AI is not just ‘predictions’ also its mainstream now bcos it is available to masses freely or at cheaper monthly subscription. Many organisations must already be using it for specific purposes and not just for ‘chating’ to get information.

github copilot we use is learning itself. Over few weeks it could understand how I name variables or functions. And even predicted what would be my next method in a class! Coding and writing unit cases was never so easy.

just give name of the class and ask it to write unit test cases. Almost all 5 service classes gets unit test ready in 10-15 min (after fixing some errors obviously)

[SchumiStars]

And then why would you want AI doing the work for you?

I actually enjoy thinking, designing, writing code and the tests for it. Unless there is a requirement for me to use AI or specific automation tools, I would prefer to use my own resources to understand and solve the requirement.

As I said previously, I like development work as it represents a challenge, sadly it does seem it's getting much easier.

Infact, what you are doing by making it easier for yourself is that you are deskilling and removing the thinking process required for getting to the result, instead of finding a shortcut to the solution.

And then what about interview questions and tests? Are you really going to say, I sat back and all my tests were created for me?

It does remind me of the days of MS front-page to create webpages. And we had people saying that anyone could write and build websites at the time. 25yrs latter whilst it is possible now, having a data-centric site still requires a considerable amount of designndev work and testing.

Even with all the tired and tested training patterns we have, there are no short cuts to running a marathon. You have to run every inch. There are no short-cuts.

Software development is a marathon, whichever way you cut it.

[willendure]

If AI was that exiciting we would have seen something by now. Unfortunately, as I thought, it was a very fast fad, which as come and gone.

AI, crypto are fads which major investment has not got behind. If they had done, early, we would not be in this hole.

The banking sector is key. If they get behind something then everyone else will. Until then, there will be no investment into IT and contracting will be pushed into another supply and demand by consultancies.

IR35, Brexit, COVID, large migration, WFH, No investment. At some point, the market will die, which is what we have at the moment.

I would love to be wrong. I would also love to work from home on a big rate but can accept that this will also need to change.

Gosh I miss going to the office.

Are you kidding? I can see why you are struggling if you think like that. I always think as contractors in a very technical realm, we have to keep up with new advances to stay current, to be able to charge a premium for our services.

Yes, there are definitely financial constraints right now dampening the market.

I think AI has been slow to get going for contractors, and reasons why that might be the case are in the "AI Surging" thread. I have been monitoring jobserve ads for AI over the last year or more, and I have definitely noticed a pick up in the number more recently - over the last 3 months or so. Sorry, no actual data - maybe I could dump my RSS feed into an AI and have it analyze it. I do feel confident now that we will see more AI contract work, and that will almost certainly be the case if the overall market improves.

I am using my time on the bench to build out my Home Lab machine in a server case I have been running for the last 10 years or so. Finally got around to installing a GPU, and spending a lot of my time getting deeper into AI. python, Keras, Max/Mojo, MlFlow, Ollama, LangChain, etc. Its a large field and there are so many tools to learn. Partly why I am monitoring all AI job ads to build up an understanding of what tools are commonly used and might be in demand and what things will be essential skills to have.

I don't think AI is the only route to finding work, but I think its a very hard domain to get into because of the mathematical skill set required. So I see opportunity there, if I can master things that other people cannot.

[willendure]

And then why would you want AI doing the work for you?

I used it today to prepare my accounts for my AirBnB. Couldn't be arsed to do it myself. That is what AI should be doing - the work we do not want to do.

[willendure]

What about aviation or the defense sector, are they now using AI to help them?

Drones, Anduril. Automation generally, guidance systems, the defense applications are huge (and well funded).

How about sport, has AI managed to work out who will win the premier league, it certainly does not look like Arsenal and if so, how much money should I bet? It's AI, right, it can't be wrong.

Used by Liverpool apparently, to gain an edge in the transfers market leading up to Klopps run of success.

What about the investment banks, can they save loads of money or are the stockbrokers now using this new tech to find new avenues to invest in?

You bet they are! Investment banks are the first customers of Extropics cooled to minus 200+whatever Aluminium chip directly running energy based model. Quants have always been into AI.

Outside of ChatGBT and CoPilot, AI is deader than dead. Anyone looking or banking on that will have a hard awaking.

Right...

There will be a bursting of the AI financial bubble for sure. But think about the internet and the dotcom bubble - in 2001 the promise of the Internet had not yet come to be, but it was massively hyped AND already priced into the inflated market prices of .com companies.

Somewhat similar situation to today - there will be an AI driven future with fusion power and flying cars and talking robots and all that. Its hyped into a price bubble which is unsustainable. But just as with the internet post 2001, the real work will continue, and continue to grow as a sector.

[SchumiStars]

Apparently the end vendor only wanted someone from a list of European countries. Comforting to see it isn't just the UK that has stopped knowing what it is doing

I used it today to prepare my accounts for my AirBnB. Couldn't be arsed to do it myself. That is what AI should be doing - the work we do not want to do.

And how did you know the results were as expected?