Using ChatGPT to write comments/posts about dive equipment

[kimh]

I spent half the thread saying "AI" has been poorly defined.

It's useless to argue about what "intelligence" is, when there's no dialog, much less agreement, on terminology.

I'm reposting the chart below – an AI taxonomy, with 3 levels, & 9 different "flavors" of "AI" functions. It's one person's view from a year ago, so it's only a jumping-off point.

But- a useful discussion about current "AI" products would relate to taxonomies like this for context, including what directions individual products are moving in, and how fast.

Instead, we get simple-minded slogans, like "marketing babble and party tricks." As I said earlier, 2023 will be a year of discovery for many — for all of us, really. But- some more than others.

 

[SlugLife]

I don't think the they'll get there any time soon. The self driving, the translation, the auto research, the pictures, etc. they all seem run into the same problem.

I wasn't predicting timelines, just that it's solvable. (Coaching from the bleachers) I think the major problem with self-driving is they're too focused on using the same hammer of Machine Learning & Machine-Learning Experts. & Image Analysis. That gets them incrementally closer. The breakthrough will come from somewhere else.

Making it look good and getting to 90% by automating the basics is one thing but the real problem lies in the last 5% or 10%. To make it great/reliable you seem to need actual reasoning, judgement or understanding or whatever you wanna call it.

Absolutely correct. The breakthrough I mention above would likely (at best) move the bar from something like 90% to 98%. Though, I'd also speculate that 8% would introduce all kinds of new or unanticipated problems.

For example, I've speculated that a "Self driving only highway lane" where all cars reported exact positioning and sensors could detect anomalies in the road, could be a semi-realistic (if very expensive) way to shortcut going from 90 to 98%. That said, what happens when a deer crosses the road, an item falls off a truck, there's an oil or gravel spill, etc?

That last "10%" is always, always, always the big one. IMO, anybody who think they can solve that last 10% all at once in any field is beyond naïve. At best, you might be able to solve 80% of that remaining 10% ... while praying to the gods that you don't immediately double or triple the number of problems or complexity.

 

[berndo]

Instead, we get simple-minded slogans, like "marketing babble and party tricks.

You posted really bad marketing text about the apple watch an called it 'pretty impressive', sh_tty pictures and a pretty bad article that's supposed to make people panic over a chatbot. Sorry, not impressed by your little presentation.

 

[kimh]

Some possible paths to AGI. Detailed, but not overly technical. No linear algebra!

 

[kimh]

Amazing images/videos. This one is just fun:

 

[kimh]

You posted really bad marketing text about the apple watch an called it 'pretty impressive', sh_tty pictures and a pretty bad article that's supposed to make people panic over a chatbot. Sorry, not impressed by your little presentation.

Look at my taxonomy again, and don't worry as much about changing the subject.

Hope this helps!

 

[berndo]

For example, I've speculated that a "Self driving only highway lane" where all cars reported exact positioning and sensors could detect anomalies in the road, could be a semi-realistic (if very expensive) way to shortcut going from 90 to 98%.

Highway driving might already be safer than human driving on average. What I call BS on is that you can get in you car in NYC, tell the car a pittsburgh address and the car just drops you off there.
That's what was promised to be possible by 2017 along with AI robo taxis.
Until I actually see a chatbot producing somewhat accurate and reliable research on some basic stuff, I'm not buying all the stories and BS about "the scared experts".

 

[berndo]

Unless something unpleasant happens with humankind in the next few years, I imagine CCR will eventually become the norm, usable by any diver.

It would require compact AI hardware, with an extensive sensor suite that tracks all relevant metrics, incl. biometrics. Since the redundant dive computer has human-level intelligence (at minimum), it can safely monitor everything, and inform/warn the diver about any issues, and provide complete, conversational guidance. Sort of like Iron Man w/Jarvis. But you’ll still have to wait for the relevant PADI course.

Just in case somebody missed this in the other thread... AI hardware... if human kind is still around in a few years.

 

[kimh]

Just in case somebody missed this in the other thread... AI hardware... if human kind is still around in a few years.

Much smarter technologists than you or I – aren't laughing about this. Doesn't mean it's likely – in fact, most of these technologists are betting against it.

But they're not laughing, either.

 

[L13]

Much smarter technologists than you or I aren't laughing about this.

Berndo is the smartest most knowledgeable technologist ever. If he laughs, it must be so. He is never wrong.