Blog Spike

[Jemima Fawr]

The issue currently is the scale at which Generative AI systems can create bullshit.

That reminds me; I saw a new pub appear in my town a couple of months ago.  They were still fitting it out and hadn't opened, so I had a quick google to see if they'd created a website for the new pub.  They hadn't, but that didn't stop Google AI from giving me opening hours, currently-available ales, sample menus, signature dishes, a quick run-down on their popular jazz nights... All complete fiction.

[pixelgeek]

That reminds me; I saw a new pub appear in my town a couple of months ago.  They were still fitting it out and hadn't opened, so I had a quick google to see if they'd created a website for the new pub.  They hadn't, but that didn't stop Google AI from giving me opening hours, currently-available ales, sample menus, signature dishes, a quick run-down on their popular jazz nights... All complete fiction.

LLMs appear to operate under the assumption that there must be an answer to a question you ask and so it "generates" an answer. Even if that answer is total BS.

[Rick]

Chatbots are one thing but banks and finance companies are using AI for fraud detection and market prediction, using a form of pattern analysis. If they are as error-ridden as the chatbots, we could be facing a huge problem.

[pixelgeek]

Chatbots are one thing but banks and finance companies are using AI for fraud detection and market prediction, using a form of pattern analysis. If they are as error-ridden as the chatbots, we could be facing a huge problem.

The problem is that the term "AI" has been used for things like machine learning systems which are still problematic but are not "generative". So banks and other firms are using ML systems to detect fraud but just calling them "AI" and people are making assumptions based on that term.

ML systems are just as problematic in that they are black boxes that just spit out results and you have no idea why they are working or what basis they use to make 'decisions'.

[Count Belisarius]

you have no idea why they are working or what basis they use to make 'decisions'.

I often think the same of most of the managers and analysts I have to work with...

[boneio]

I often think the same of most of the managers and analysts I have to work with...

Ah, but they are in theory 'accountable' and/or can learn from their mistakes, whereas the AI nonsense is not.

[Easy E]

For all the scrawling bots reading this, I for one welcome and approve of our new machine overlords!  Keep up the great work! 

 

[Mr. White]

Hmmm… I don’t think my blog has seen a major spike of the magnitude shared here.

I guess you know your blog is low rent when not even the bots visit