AI fails

[Psamathe]

Re: Data Centres
Those in favour of these massive energy & water guzzling developments often use the argument we need them as we need to stay in the AI game and if we don't they'll be in the US and we will be dependent. (An argument Rory Stewart keeps using).

The thing I think they are missing is that these developments will be paid for by US money, will contain US chips and run US software which means Trümp and subsequent US Governments and commercial companies will have the "kill switch" so we will need to be keeping them happy or they'll pull the plug (within 1 hour).

So we decide to regulate Social Media to protect our children from life impacting damage or decide to tax those companies and suddenly the AI we thought we had, we haven't and without notice it's gone.

 

[Pblakeney]

Re: Data Centres
Those in favour of these massive energy & water guzzling developments often use the argument we need them as we need to stay in the AI game and if we don't they'll be in the US and we will be dependent. (An argument Rory Stewart keeps using).

The thing I think they are missing is that these developments will be paid for by US money, will contain US chips and run US software which means Trümp and subsequent US Governments and commercial companies will have the "kill switch" so we will need to be keeping them happy or they'll pull the plug (within 1 hour).

So we decide to regulate Social Media to protect our children from life impacting damage or decide to tax those companies and suddenly the AI we thought we had, we haven't and without notice it's gone.

Not only that but the US is kicking back so they are merely outsourcing the problem.

 

[Psamathe]

Not only that but the US is kicking back so they are merely outsourcing the problem.

Another case of offshoring your pollution?

 

[Pblakeney]

Another case of offshoring your pollution?

Yup!

 

[Ian H]

I'm not sure whether this counts as AI success (for Uber & co) or failure.
Food delivery riders call on platforms to open up AI ‘black box’ they say has cut pay
Workers who blame algorithms for lowering their earnings are getting help from academics to find out how the system works
https://www.theguardian.com/busines...ry-riders-platforms-open-ai-black-box-cut-pay

 

[briantrumpet]

An interesting piece from a coder who embraced AI to help him code until he realised he had to stop.

https://brettcodes.com/im-done-using-ai/

 

[C R]

An interesting piece from a coder who embraced AI to help him code until he realised he had to stop.

https://brettcodes.com/im-done-using-ai/

View attachment 17089

That's my feeling on it too. Our company had an initial big push on using it. Some of us were skeptical, partly for the reasons above, and some were quite enthusiastic. The enthusiasm has slowly given way to the realisation that all AI does is type faster, but what it comes up with is often not very good, and needs so much handholding that often it is quicker to just do the job yourself.

 

[briantrumpet]

That's my feeling on it too. Our company had an initial big push on using it. Some of us were skeptical, partly for the reasons above, and some were quite enthusiastic. The enthusiasm has slowly given way to the realisation that all AI does is type faster, but what it comes up with is often not very good, and needs so much handholding that often it is quicker to just do the job yourself.

Do you think it will get much better, or is it inherently problematic?

 

[C R]

Do you think it will get much better, or is it inherently problematic?

For the mechanical part of the job, where you can define exactly what you want and there are no external dependencies, it probably is good enough already. The difficulty is when it comes to general design based on user requirements. User requirements are rarely captured in sufficient detail, we are some way away from AI to do a good job.

Edited for spelling.

 

[Psamathe]

That's my feeling on it too. Our company had an initial big push on using it. Some of us were skeptical, partly for the reasons above, and some were quite enthusiastic. The enthusiasm has slowly given way to the realisation that all AI does is type faster, but what it comes up with is often not very good, and needs so much handholding that often it is quicker to just do the job yourself.

I remember many years ago when some larger companies tended to make a lot of use of software development from one particular overseas company (because they were cheap) and we did one project where we were developing the mobile side of the project, developers from said country the Head Office systems and ... they were very good at designing database structures, entity relationship diagrams, etc. the internals but what they produced was a disaster because they could not understand (didn't even look at) how the users worked, how they interacted with the systems and what they needed to do with the system.

Made me wonder if AI might suffer similar? (Not used AI so not suggesting this is the case).

 

[C R]

I remember many years ago when some larger companies tended to make a lot of use of software development from one particular overseas company (because they were cheap) and we did one project where we were developing the mobile side of the project, developers from said country the Head Office systems and ... they were very good at designing database structures, entity relationship diagrams, etc. the internals but what they produced was a disaster because they could not understand (didn't even look at) how the users worked, how they interacted with the systems and what they needed to do with the system.

Made me wonder if AI might suffer similar? (Not used AI so not suggesting this is the case).

Yes, that's a good comparison. I also worked at a place that 'enjoyed" the privilege of such outsourcing, and the feeling is the same after you spend two weeks refactoring something you could have done in two days.

It may be different in green field projects, but in a mature situation whereby there's a significant amount of info that is not explicitly recorded anywhere, as well as technical debt, AI gets the wrong end of the stick more often than not.

 

[Pross]

We’re starting to get clients who are emailing obvious AI answers arguing against advice we give that they don’t like or responses from planning officers. It’s so obviously come out of AI and completely lacks the nuance that the not highly enough paid experts are giving them. The worst offender though is one that takes our advice then runs it through AI before emailing it to planning officers rather than just letting us submit directly.

 

[Pblakeney]

It is not just luddites and doomsayers warning about the rise of AI.

"OpenAI's chief scientist Jakub Pachocki has called for "extreme caution" over AI's runaway progress and warned more intervention may be needed to ensure "humans remain in control of the future".
"I am concerned no one is prepared for the consequences of a continued rapid rise in machine intelligence," he wrote in a blog post entitled "An Alien Mind, external"."

 

[farfromtheland]

G-corp Deepmind 4, Lee Sedol 1. That was the seminal moment.

Later I had some fun reading an AI dialogue about Noam Chomsky's views on AI.
I don't know what exactly comprises 'soul' but am completely convinced that the lack of it is a problem.

 

[Ian H]

G-corp Deepmind 4, Lee Sedol 1. That was the seminal moment.

Later I had some fun reading an AI dialogue about Noam Chomsky's views on AI.
I don't know what exactly comprises 'soul' but am completely convinced that the lack of it is a problem.

A problem just for AI, or for humans as well?