All things AI.

grunge said:
My hunch is, if the AI bubble is going to burst, its going to burst very soon. Looks to me as though the Venture Capitalist money is starting to dry up and AI companies are starting to realign costs and pricing to remove subsidised pricing.

Anthropic have changed what you get in your subscriptions. the top subscription they have is $200 a month. I'm reading lots about the fact it used to last people the whole month or at least a good few weeks is now being burned through in 2 days, there pricing all round has just sky rocketed.

Github Copilot is moving from a subscription model to a Pay as you go model as of June 1st. I can see Microsoft doing this for standard copilot in windows next.

We are in a situation where 10's of thousands of people have been laid off for the move to AI as CEO's saw it was cheaper but looks like its could be cheaper to actually hire people back than us AI to its fullest...

we may just see a run on RAM again as companies try to move to inhouse services and have to buy the hardware.
Click to expand...
With the money sloshing around in an effort for “Big Tech” to own the AI infrastructure, corner the markets of both chips and electricity generation, and layoff as much high priced “yesterday” talent as possible, we are approaching an inflection point where the $Trillions in investments is either going to pay off and inexorably change the global landscape OR people will realize there are much cheaper work arounds with almost identical, if not identical, results.

In fact, I think this Xi-Trump Summit will start to lay out some of the hi- vs lo- tech vision of the future, based on lowest cost commodity compute, and who is building it!

Personally, I’m hoping I can reasonably time that inflection point and save my investments before the explosion happens and trillions are wiped of the balance sheets of punters like us.

Jensen can only try to outsmart his audiences for so long…
 
I'm an advocate of using AI and have inputted into a fair use policy around how as an organisation we should embrace it ethically. I use it as part of my job. What people don't always understand is, AI still needs human interaction. We can't just take everything it spews out as gospel. It will hallucinate and go off in its own direction.
We use it to help us cut down on mundane or highly time consuming activities eg one of the things I'm currently looking at for using an AI solution - it takes around 90 hours for 1 person to complete a certain activity, we tested this using an agent we built and it took less than 5 minutes to deliver the first report but after a few interactions and training with our resource, we completed the task in 1 hour. Things like this allows our specialist resource to focus on more critical activities.

As others have said, AI needs people and the move to more agentic AI will need specially skilled people to create agents and requirement for prompt engineers etc.
 
crystal_mais said:
I'm an advocate of using AI and have inputted into a fair use policy around how as an organisation we should embrace it ethically. I use it as part of my job. What people don't always understand is, AI still needs human interaction. We can't just take everything it spews out as gospel. It will hallucinate and go off in its own direction.
We use it to help us cut down on mundane or highly time consuming activities eg one of the things I'm currently looking at for using an AI solution - it takes around 90 hours for 1 person to complete a certain activity, we tested this using an agent we built and it took less than 5 minutes to deliver the first report but after a few interactions and training with our resource, we completed the task in 1 hour. Things like this allows our specialist resource to focus on more critical activities.

As others have said, AI needs people and the move to more agentic AI will need specially skilled people to create agents and requirement for prompt engineers etc.
Click to expand...
Or..you now only need to employ one ninetieth of your specialist resource team
 
citizen_maine said:
Or..you now only need to employ one ninetieth of your specialist resource team
Click to expand...
For the work we are doing, we have a backlog of around 20 of these activities. so while he/she is doing 1 there are about 5/10 other activities we need them to do and the backlog of another 19. We actually have 5 of these resources currently.

in that 90 hours on just 1 design - things will/could change that would mean some of the work then becomes un-useable. Due to backlogs today we use a very high level methodology which is less manual and only takes about 2 days effort but the level of risk is higher as we may miss things.
 
crystal_mais said:
For the work we are doing, we have a backlog of around 20 of these activities. so while he/she is doing 1 there are about 5/10 other activities we need them to do and the backlog of another 19. We actually have 5 of these resources currently.

in that 90 hours on just 1 design - things will/could change that would mean some of the work then becomes un-useable. Due to backlogs today we use a very high level methodology which is less manual and only takes about 2 days effort but the level of risk is higher as we may miss things.
Click to expand...
Yeah, I'm just cynical mate having spent over 40 years listening to 'Yeah, but we'll be able to do all those jobs we're supposed to do but haven't had the time/staff to do'.
What then happens is somebody leaves (either through choice or redundancy) and everyone else picks up the slack and you're back to where you were, but the employer is paying fewer salaries
 
Ai used in the right way would be the most amazing development ever seen, it's potential is jaw dropping.
The trouble with the world we live in is it will be highjacked for power, control and greed.
 
abu13 said:
Ai used in the right way would be the most amazing development ever seen, it's potential is jaw dropping.
The trouble with the world we live in is it will be highjacked for power, control and greed.
Click to expand...
Anecdotal, but probably not only happening to me - I have only recently started using AI for coding (Codex within VSCode).

The reason is that my client specifically wants their web app re-written in React, which I have never used.

It's incredible how Codex writes the code, refactors (if ChatGPT doesn't like how it's structured things) where necessary, etc without me even needing to copy&paste - but I have learned nothing of how to develop in React in the meantime (which is fine in my case, as long as it can do the job)

The problem is, without any new coders actually writing human-developed code, won't AI simply be using its own back-catalogue to seed its own answers going forward?

Not a great model for progress, that.
 
citizen_maine said:
Yeah, I'm just cynical mate having spent over 40 years listening to 'Yeah, but we'll be able to do all those jobs we're supposed to do but haven't had the time/staff to do'.
What then happens is somebody leaves (either through choice or redundancy) and everyone else picks up the slack and you're back to where you were, but the employer is paying fewer salaries
Click to expand...
I know mate - we are actually employing a few more specialist as it will take at least another year for us to find the right solution. My view is for people to upskill with AI engineering skills too. AI done right adds value, AI done incorrectly causes lots of errors - It's only as good as the data it has access to and the person using it properly with the knowledge of what it's trying to ask it to do.

I can see some orgs wanting to use it to reduce headcount - good luck as we then go into a world of potential incorrect use. There are programmers out there asking AI to write code for them and then productionising it without audit trails :(
 
With regards to costs. Microsoft have blocked engineers using Claude and are apparently terminating the contact.

Uber burned there whole year’s worth of AI budget in 4 months.

www.forbes.com

Uber Burns Its 2026 AI Budget In Four Months On Claude Code

Uber exhausted its 2026 AI budget in four months on Claude Code, exposing how token pricing breaks enterprise finance assumptions.
www.forbes.com www.forbes.com
 
Long story short, I just started using AI 3 months ago to distribute my own creations. Each creation viewed gets me some earning.

My current passive income is US$1,200 per month, which I believe will increase month on month.

I allocated mine to the US demographic because there are more audience there but at the same time also has global viewership.

My creations are nothing inflammatory, no politics, no explicit content, no humour. In fact I’m certain mine’s for all ages. It’s just creativity.
 
I am afraid, too many employers are relying on AI, and when the Chinese turn off the internet we will no longer be able to access banks, savings, business will close down and we will be back to the dark ages.
 
Bigga said:
The AI bubble is close...

Click to expand...


We might be able to avoid a huge bubble burst for a couple of reasons neither of which are particularly healthy: firstly the really big players like alphabet, microsoft etc are so big they might be able to weather the storm. Secondly the legions of overvalued startups will try and recalibrate and pivot to different and more specific business models that are less about their LLM etc and more about what else they've learnt along the way, no doubt some of it nefarious! The percentage and degree to which they are successful might determine how big the bang is in reality. That said plenty of scummy but smart people seem to have exited nvidia of late so who knows.
 
Bigga said:
The AI bubble is close...

Click to expand...


Bit of an over simplified headline that one. they stopped engineers using Claude from Anthropic due to excessive costs. Uber used there entire years worth of AI budget in 4 months because of it. they have limited there engineers to $1500 of tokens a month... thats still A LOT.

Yesterday MS announced there own frontier LLM model that will probably start to ramp up usage internally before they sell it on.

MS have however just changed Github Copilot pricing and have cut the usage down by extreme amounts. I managed to use 90% of my usage allowance in 2 days, I used to be able to use it for the whole month with no issues.
 
Quite noticable at work that AI is being pushed in a big way, all the management team are stressing the importance if it and the opportunities it offers. Right up until anyone mentions the impact on job markets and then we get cut off.
Many of us are now in a race to make ourselves dispendable.
 
threespires said:
We might be able to avoid a huge bubble burst for a couple of reasons neither of which are particularly healthy: firstly the really big players like alphabet, microsoft etc are so big they might be able to weather the storm. Secondly the legions of overvalued startups will try and recalibrate and pivot to different and more specific business models that are less about their LLM etc and more about what else they've learnt along the way, no doubt some of it nefarious! The percentage and degree to which they are successful might determine how big the bang is in reality. That said plenty of scummy but smart people seem to have exited nvidia of late so who knows.
Click to expand...

I'm interested in how you're getting to your thought process.

At what point are you starting the equation?
 
You must log in or register to reply here.
Unread Posts

Don't have an account? Register now and see fewer ads!

SIGN UP
Back
Top