Will AI Kill This Group?

That's the catch. AI is just the collective wisdom of a large collection of text, processed by a large language model, and Cursitor Doom's contributions get just the same weight as those from Phil Hobbs.

Dealing with false information may look like a pissing contest to Curistor Doom, but he's no more capable of distinguishing fact from misinformation than any large language model.

Except that it isn't super-intelligent. It isn't intelligent at all. It's just very well-informed, except that it can't weigh the reliablity of the information it retails.

That depends on how many of the participants are as ill-informed as Cursitor Doom.

The gullible goofs may prefer AI, in the same way that Cursitor Doom prefers to get his news from Russia Today, but Usenet is a forum for human interaction - sub-human when Cursitor Doom is involved - and AI doesn't seem to be able to offer that yet.

Bill Sloman's comments about Cursitor Doom and John Larkin could have been written by a machine - but they don't appear to be intelligent.

Indeed. Bill illustrates my point without intending to, in fact. One doesn't get a mouthful of abuse from AI merely for asking a straightforward question. The main problem with Usenet AISI is that you can get a response from say Phil Hobbs on the one hand, but you might equally well end up with one from Bill Sloman. Still, at least with Usenet, you get a killfile to deal with the trolls, so that's one plus I guess.

You have to understand what you are reading before you can make a worthwhile comment about it's intelligence. Do find a machine that can match my idiosyncracies - you may find it a little more difficult than you imagine.

Nor is that what you get from me - at least not in response to a straightforward question.

You occasionally get one from us both, if you ask a particularly good question - not your forte.

Even trolls have their uses. Goofs like you occasionally ask peculiarly ill-informed question, and can be abused for it without there being any risk that the response will put them off asking less half-witted questions. Dim newbies get more gentle treatment than equally dim case-hardened trolls.

In that case I wouldn't use it yet because there's too much risk that it's inaccurate.

Ask it to design a low distortion sinewave oscillator and it won't come close to what the recent collaborative effort here produced.

Not yet.

There's also the issue of web sites produced by bots which were trained on web site content.

Perhaps you should train one. Then see how many people want to use it. There will of course be some people who do want to use it because they will be interested to see what kind of insults it comes up with.

First I'd have to get my hands on one to train, then I'd need to acquire the relevant skills. I actually know a few people who have those kinds of skills, but only because of a friendship that goes back to the

The insults will all be variants on "you don't know what you are talking about" which is - of course - insulting, even if it is a bald statement of fact. It's not a particularly entertaining kind of abuse, even if you put in some effort into getting some variation into the message.

It might be able to extract a set of the questions we end up asking dim newbies when they come here for help, but it isn't obvious that there's a single one-size-fits-all set of questions.

"What exactly are you trying to do", and "what made you think that that particular approach might work" do come to mind, but they aren't diplomatic enough to be all that useful.

You only have to verify what it comes up with. Don't throw the baby out with the bath water as you'll be seriously missing out.

Yes, I remember that. Massive credit to you for coming up with those designs with Bill's constant interference to contend with.

It's coming on in leaps and bounds, though; getting stronger and stronger as time goes by. It pays to check it out periodically as it becomes increasingly refined.

You have friends?? Great god alive!

There's one thing that AI will never be any use for IMO and that's making qualitative judgments about music. I recently had a blazing argument with Grok over who was the better guitarist out of Jimmy Hendrix and Hank Marvin. Grok insisted that Hendrix was not only better, but the best guitarist of all time! I pointed out that without the input of some very clever British sound engineers and their early SFX, Hendrix would never have made it. But Grok was having none of it! I thought it was a bit much for a machine with no ears to be making such judgments.

With some types of 'music' that gives it an unfair advantage.

Sure, and I've been horrified by what it did. It's true that that was a long time ago.

Sure but I don't recall it catching anything I couldn't have caught myself.

CAD in various forms has been part of electronics since I started work. At that time I was lucky to work for a manager who knew the limitations of CAD. No doubt things have got much better now and will get even better in the future. I would no longer design any kind of complex sequential digital circuit without VHDL or Verilog. But

Interestingly, ORAC does seem to have personality traits similar to Bill.

Why are newbies dim Bill? Few people come here asking for help and I don't think it's surprising. You mention Phil Hobbs a lot for some reason. I've never seen Phil insult anyone and would not expect to do so.

Checking the netlist is essential. Why is there a separate gnd and grnd or ck and clk net? Maybe it was intentional but often not. Not to mention the board which was expected to work without any of the chips connected to any power supply or ground at all.

I noticed an AI response recently which implied that op amps require a +15V and -15V rail. And one which said an op amp is an integrated circuit.

I'm good at doing what others think are menial tasks, and uncovering issues which might otherwise have needed expensive redesigns.

No but I do have a feel for it so that I can check what should be checked. I recall a power supply which would otherwise have had many amps through one tiny through hole.

True, but a CAD tool won't say "Excuse me are you sure you meant to have separate gnd and grnd? At least not one I've ever used. No amount of looking at the schematic will reveal issues like this. But a quick look through the netlist will.

Sure but why didn't the CAD tool say "Excuse me you moron your chips have no power supplies" Or something like that.

Try asking how many t's there are in Stuttgart.

At some things. Not so good at others.

If you don't know what should be designed for current carrying capacity then you shouldn't be designing it in the first place.

I usually type zzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzz zzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzz zzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzz zzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzz And then ask them to fix the unchecked buffer.

Validating input is the first rule of um well anything which accepts input.

Your grasp of reality has always been feeble.

By no means all newbies are dim. The ones that ask ill-formed questions often are. That question wasn't exactly well-formed.

He's a consultant. He doesn't want to put off potential customers.

John May contributed rather more than Edward Rawde, and John and I managed to get rid of a few the transistors that Edward chose to introduce.

It's getting the benefit of even larger large language models. The basic idea hasn't moved forward, and there's no sign that anybody has incorporated anything like critical thinking into the process.

That might be the next step up, but it would call for a very different kind of processing.

You don't know what "critical thinking" implies, so it isn't surprising that you've missed that point.