On AI mania

[Bodhitree]

I just came across this essay on AI mania in the business community, which I found very interesting. It’s like investors have found it so enticing, that it is being forced on everyone from board of directors down to CEOs, executives and employees, without real proof that it is effective.

It seems to have become a fervour in an almost religious style, where people who don’t believe end up getting fired, and all kinds of inefficient behaviour is being encouraged to keep the push for AI in the workplace intact. The article goes into various types of behaviour at various levels.

AI Mania Is Eviscerating Global Decision-Making — Ludicity

What do you all think? Do these claims of massively increased productivity actually hold water, or is it like the writer of the essay says, largely overblown sales hype?

 

[Scepticalscribe]

The latter, - very much the latter - in my view.

 

[Bodhitree]

It certainly seems that way… as the first article stated, a 10% error rate may be acceptable when searching for a peanut sauce recipe, but imagine your CFO getting the wrong number one time in ten when compiling your financial year report.

According to The Guardian the jobs market certainly doesn’t seem to be transforming as quickly as some had predicted.

The AI jobs apocalypse probably isn’t coming anytime soon

Artificial intelligence may not deliver on its promise of vast economic opportunity at a price that humanity is willing to pay

www.theguardian.com

 

[KaliYoni]

At the risk of sounding glib, I'd say Suresh has "meme envy" plus something he's trying to sell. If we were in 2017, not halfway to 2027, he'd be sounding a similar alarm about blockchain.
🙂

‘A feedback loop with no brake’: how an AI doomsday report shook US markets

Shares in Uber, Mastercard and American Express fall on back of apocalypse scenario posted on Substack

www.theguardian.com

Skittish investors spooked as dystopian AI outlooks go viral

A recent huge bet on the technology is starting to show cracks.

www.reuters.com

 

[Bodhitree]

I think he sounds like he is trying to explain what he is seeing, this crazy behaviour of AI at any cost. I think Apple’s take, of gradually introducing the technology, is the state of the art in actual working implementation.

Suresh is operating from a limited view of business, and he may be overly pessimistic about where this will go in say 5 years. But certainly right now, the term ‘AI mania’ seems appropriate.

 

[throAU]

What do you all think? Do these claims of massively increased productivity actually hold water, or is it like the writer of the essay says, largely overblown sales hype?

The truth is somewhere in the middle

• There's a huge amount of hype
• The higher end models have massive capabilities when used correctly.

For my work:

• LLMs make finding and filtering information from the web much faster
• LLMs can write proof of concept code, or single use code. much, much faster to get work done
  • This means testing an idea is much easier, and you're far less tied to the code due to the sunk cost fallacy. Instead of being multiple days/months/weeks to test an idea in code, its minutes or hours. If the idea doesn't work, the project can be thrown out, rather than becoming attached to it and trying to make it work due to the sunk cost in labour.

On the flip side

• LLMs are still massively prone to hallucinations, and output NEEDS to be vetted by a human

Think of them more as a tool like a hammer or a screwdriver and less as a human replacement and they're and extremely powerful force multiplier.

Companies getting rid of humans for AI are missing the point. AI companies themselves are on a hiring spree, if current AI could replace humans they would not be doing this.

 

[wonderings]

We are looking at ways of utilizing AI to our benefit, not for the sake of just using AI. It is a helpful tool for various tasks but it is still used randomly for different things, it has not taken over one task or job and replied it completely with AI. I got my staff ChatGPT Pro and I do go over how it can be useful and helpful. It is definitely a learning thing for people to understand there are some really amazing things it can do. Two quick use cases where AI sped things up considerably.

1) A large client needed some vinyl decals printed in a rush for an event happening this weekend and they needed them for this afternoon to setup. The logo they supplied (from a sponsor of theirs) was really bad, very pixelated and they had no access to anything better. I dropped the logo in ChatGPT and ask for a cleaned up but as accurate as possible recreation of the logo. In under a minute I had a high enough res version that looked a thousand times better then their file and was able to make a decent print ready file. No working with the logo, no special handling, just AI and a command. Of course this logo was more on the simple side geometry wise, this will not work with everything. But this is where the education comes in on when and how it can be useful.

2) A CSR had a strange print request, something we would have to outsource but have never done before. CSR was not even sure how to properly word her google search to try and find something. I told her to ask ChatGPT and again in seconds we had a supplier less then an hour away that did exactly what was needed. This replaces google as. your prompt can be more vague and it can help suss it out.

Two very different tasks that saved a lot of time. What I am working on educating my staff with is understanding when to use it and when it can be helpful. That basic understanding makes it a very valuable tool. So I do think it is necessary and with new hires we have asked about their comfort with AI. It is not at the point where we would not hire someone who did not have knowledge with it but it very well could be sometime soon.

 

[throAU]

We are looking at ways of utilizing AI to our benefit, not for the sake of just using AI. It is a helpful tool for various tasks but it is still used randomly for different things, it has not taken over one task or job and replied it completely with AI. I got my staff ChatGPT Pro and I do go over how it can be useful and helpful. It is definitely a learning thing for people to understand there are some really amazing things it can do. Two quick use cases where AI sped things up considerably.

1) A large client needed some vinyl decals printed in a rush for an event happening this weekend and they needed them for this afternoon to setup. The logo they supplied (from a sponsor of theirs) was really bad, very pixelated and they had no access to anything better. I dropped the logo in ChatGPT and ask for a cleaned up but as accurate as possible recreation of the logo. In under a minute I had a high enough res version that looked a thousand times better then their file and was able to make a decent print ready file. No working with the logo, no special handling, just AI and a command. Of course this logo was more on the simple side geometry wise, this will not work with everything. But this is where the education comes in on when and how it can be useful.

2) A CSR had a strange print request, something we would have to outsource but have never done before. CSE was not even sure how to properly word her google search to try and find something. I told her to ask ChatGPT and again in seconds we had a supplier less then an hour away that did exactly what was needed. This replaces google as. your prompt can be more vague and it can help suss it out.

Two very different tasks that saved a lot of time. What I am working on educating my staff with is understanding when to use it and when it can be helpful. That basic understanding makes it a very valuable tool. So I do think it is necessary and with new hires we have asked about their comfort with AI. It is not at the point where we would not hire someone who did not have knowledge with it but it very well could be sometime soon.

If you have creative/imaginative users, a ~$30 AI subscription will pay for itself in saved time in a single automated task per month.

I regularly use LLMs to assist with change request creation, instead of me writing a bunch of stuff to summarise a set of steps into change management speak, the LLM can do in 30-60 seconds what would take me 30-60 minutes to write. I need maybe 5 minutes to give it a proof read.

I do multiple change requests per week
my hourly rate is way more than the monthly sub cost

 

[wonderings]

If you have creative/imaginative users, a ~$30 AI subscription will pay for itself in saved time in a single automated task per month.

I regularly use LLMs to assist with change request creation, instead of me writing a bunch of stuff to summarise a set of steps into change management speak, the LLM can do in 30-60 seconds what would take me 30-60 minutes to write. I need maybe 5 minutes to give it a proof read.

I do multiple change requests per week
my hourly rate is way more than the monthly sub cost

I have pre-ordered a device that looks very interesting and will potentially save hours. https://violoop.ai/

I will be testing it myself once they actually start shipping and see if it lives up to the hype and what is expected of it. If it does I can see easily getting half a dozen of these to start and implementing them with our sales and CSR's.

 

[MBAir2010]

I have pre-ordered a device that looks very interesting and will potentially save hours. https://violoop.ai/

Did a bunch of 5th graders design the website?
the header text is cut off on the top and the link verbiage is bad English.
then I left the site.

 

[wonderings]

Did a bunch of 5th graders design the website?
the header text is cut off on the top and the link verbiage is bad English.
then I left the site.

not a huge fan of the page but it is the device itself that interests me. Could be a dud, but if it does what is promised then I could care less what the website looks like.

Header text is not cut off for me on a 27" 5K monitor.