You are viewing a single comment's thread from:

RE: Ethical Use of AI: The Hive Perspective

in StemSocial4 days ago

A lot to take in here. One area of research that I am interested in is how AI hallucinations can create positive feedback loops, so that in further inference, AI models will double down on their "being incorrect", but confidently correct.

It is an inherent flaw in the probabilistic modelling that these models use.

When such models are also used adversarial manner ("responsible AI") or likely, in cyber sec, its going to lead to an ongoing escalation of compute more broadly as a tool.

There will always need to be a machine operator. We can't have "dark" factories of AI being autonomous, as there will need to be people with some serious critical thinking skills (and unbiased dimensions) to work within.

If we get autonomous "AI" (and the bad, generative model of AI) doing things, then that will continue to contribute to the absolute enshitification of everything as we know it.

Having said that, I can see (and indeed use) AI as a tool in my day job, day to day, and it makes me more efficient, and allows me to interrogate the datasets I work with on a deeper level than I would normally have the capacity to manually do.

As a result, it allows me to be more knowledgable of the subject matter, and to better answer business stakeholder's questions.

Simultaneously, there's a further story brewing in that field, and I'll be able to share more on HIVE about that soon when certain NDA's expire :)

Sort:  

I like alcohol, a cold lager on a hot day, perhaps a bourbon after a tough days of work. But I won't be driving after 5 bourbon, probably won't even be having 5 bourbon at my age!

That's the difference.

Also if I get caught drunk driving I got straight to jail. The law exists to protect the citizens.

Real time AI legislation wen? :)

Today, the BAC (Blood Alcohol Concentration) as a percentage is 0.05 in Australia for the line of legality, what if, during a special event it was set to 0.00, or 0.1 ?

What if it adapted based on the number of people on the roads, or on certain roads (school zones?)

There's dangerous levels of depth that AI can go down to (much like an actuary) but in the "right" hands, everyone is an actuary with the right AI (or at least, they think they are!) :)

Some regulations is better than no regulations. Thinking what if a cute bunny will do made the entire Australia a desert. You know the story I believe!

https://education.nationalgeographic.org/resource/how-european-rabbits-took-over-australia/

Yep, they're still everywhere :D