ChatGPT incoming

Yes certainly AI functions in the same way as the human brain in some areas. But the key difference is AI (so far!) has no understanding of what it is doing. It has great powers of computation, that is calculation, at great speed, good memory capacity and the ability to search the web etc and bring a lot of stuff together but zero power of comprehension. It can write an essay on how a Myford lathe compares with a Sieg lathe as shown earlier in this thread. But it has no understanding of what a lathe is and even less of why anyone would care. So far, no great threat to the status quo.

Yet some of the world's leading AI experts are concerned for the future, so there must be potential for it to develop further. And so certainly bears watching. But I see their calls for limits or restraints or rules on its development as naive.Only the good guys will follow those rules and restraints. The criminals, and rogue regimes everywhere, will carry on regardless. Better we have first and not them.

The "AI has no understanding" notion is debatable, given that words such as "intelligence" and "understanding" aren't well-defined. Really, we don't have much better than the famous definition for porn: "I know it when I see it."

That was Turing's motivation for his famous test: How can I tell if something is intelligent if I can't even define the term? He decided that if you can't tell it apart from a person (and if you assume people have intelligence), then that will do.

The reason recent AI seems magical is that we can't understand what it's doing, because the network of connections it has learned is far too vast and opaque for us to cope with. We aren't intelligent enough for that!

So instead of presuming that intelligence is a "thing" that we have to somehow find residing inside a computer, perhaps we should just ask "does it do the things that intelligence does?"

In that case, the latest AI's are at least doing things, such as writing essays and so on, that appear intelligent, e.g. hard to tell apart from what a well-educated and presumably intelligent human could do.

Edited By S K on 29/05/2023 15:42:22

Just last week a junior lawyer in the US was reprimanded by a judge. His case relied on 5 previous cases quoted in detail. He had used ChatGPT and ChatGPT had invented all of them. He had to "throw himself at the mercy of the Court". ChatGPT has previous at inventing stuff- if it can't generate an answer from its dataset, it just makes stuff up. It NEVER returns a 'don't know'.

Steve,

I guess he'll remain 'Junior' for the rest of his career doing stuff like that. I'd have thought he'd have at least done a quick check on the 5 cases to see if they were actually real. Thankfully the Judge had his wits about him.

I think ChatGPT is at the stage where it is learning how to learn....it's still basically a child.

Give it a year or two, powerful enough hardware to run on, and access to the bulk of the web and I suspect we are in for a few surprises, maybe not all good.

I can see two scenarios where it could cause humans problems...

a) it gets in to the hands of bad actors, which it will, and

b) it goes feral, which is possibly the bigger worry.

Ches

ChatGPT has also invented reference to a American Law professor being accused of sexual harrasment.
lots of references to this here is a radom one https://www.businessinsider.com/chatgpt-ai-made-up-sexual-harassment-allegations-jonathen-turley-report-2023-4?r=US&IR=T

Another issue is by entering anything into thesystem you give them the right to us it.

ChatGPT is not a tool in the typical sense.

It knows the rules of our language - how we use it and what the patterns around our questions and answers are shaped like. It then generates a 'statistically correct' response.

Within any chat you can instruct it to not return unverifiable material. It won't stop completely but it will reduce.

Precede any chat with a list of ground rules (that you develop).

These are my own observations.

Your milage may vary.

Frank C.

Just asked Bard & Chat GPT for 36mm bi metal hole saw speed in mild steel. Bard suggests 500-700 and . GPT goes for 150-200.

An internet search suggests to me 250-300

so I will need real people on an engineering forum to tell me the facts, who will also be able to add some other useful tips and not obsess about safety glasses.

An interesting [at least to me] aside:

National Savings & Investment has a very well-mannered AI ‘bot’ answering telephone enquiries

… but ‘he’ fairly quickly loses track when faced with an unusual question.

The trick to exit the endless-loop is to tell him: I need to speak to a human

MichaelG.

.

https://www.nsandi.com/contact-us

Call us free on 08085 007 007

Edited By Michael Gilligan on 15/06/2023 12:21:49

Trouble is, the 'facts' depend on a lot of variables. Asked the same question, real people will come back with different numbers and unsolicited advice too.

The RPM depends on a host of variables. Which mild-steel? How powerful is the machine? How many teeth per inch? How wide are the teeth? Is the goal to cut metal as fast as possible, or to maximise cutter life? Most recommended speeds come from industry, who go for best financial return on tool-cost, power-consumption, tool-life, and production rate. They tend to push tools much harder than Model Engineers, and would only expect a hole saw to do about 6 hours work before scrapping it. Tools last a lot longer at the rate I cut metal, but I sometimes work in a tearing hurry. Last time I cut a hole in a galvanised water tank, it was in a cramped stinking-hot midsummer loft; so unpleasant I didn't care how long the saw lasted! I wanted out, and flogged my electric drill and hole saw to get the hole cut asap.

The way questions are asked alters how they are answered. Chaps who have to ask this question on the web probably don't know about H&S either! Better safe than sorry, which is why safety glasses and other precautions are often mentioned. An AI picks up that Safety Glasses are often mentioned in answers, so it mentions them too. It's not wrong!

My answer? For almost everything, RPM = 10000/diameter in millimetres.

Smarter than it appears, the sum gives a satisfactory answer for cutting mild-steel with HSS. Other metals can be inferred from the mild-steel result, for example divide rpm by 5 for hard Cast-Iron, and multiply by 5 for Aluminium. Double, triple or even more RPM if the cutter is carbide.

Not gospel - it's a simple calculation used to get close to a speed that will work, and the operator should always expect to experiment. Having started a cut at the guesstimated formula speed, I adjust feed-rate until the machine sounds as if it's working moderately hard and/or the finish is OK. Rarely necessary to change RPM as well, but it does happen. Cutter shape, power, and machine rigidity all make a difference. If I have to look up cutting speed for something unusual, like depleted Uranium, I apply the same logic - only using the recommended cutting speed as a starting point. I always experiment to suit my equipment and needs.

My approach wouldn't satisfy a production engineer: they'd do it properly. I almost always cut slower than book speed, which would bankrupt a manufacturer! Fortunately, unless time-is-money, cutting speed is rarely critical - I find machinable metals cut reasonably well over a wide range of speeds. A lot of experimenting may be needed to cut nasty metals like work-hardening stainless steel. The best way to deal with fussy metals is to avoid them!

Dave

Though how long will it be before the formally well mannered bot, when requested to speak to a human, will snidely retort "you humans are not as smart as you think you are".

Justin asked for useful tips:

Keep the hole saw in cut and be prepared to stop to clear the kerf from time to time - Rotabroaches are better than standard holesaws for this.

For the same reason, when possible drill one or more small holes intersecting the kerf.

In a fixed machine it is not essential to use a pilot drill - useful to produce a plain slug.

Fast Mouse + Maze (25 mins) .... https://www.youtube.com/watch?v=ZMQbHMgK2rw

I have to say, those things are really fast at moving, and at calculating.

I wonder if ChatGPT, or it's offspring, will bring any improvements to this technical sport?

Ches