Aja
Well-known member
That means you don't know that you are looking at it. Nowadays AI is everywhere. I myself like to ask 'Perplexity' it's opinion on things. Haven't found an answer which I dislike yet
!
How do folks feel about Chat GPT texts on the forum?
- Thread starter fredrussell
- Start date
That means you don't know that you are looking at it. Nowadays AI is everywhere. I myself like to ask 'Perplexity' it's opinion on things. Haven't found an answer which I dislike yet
!Jerbro
Active member
An Alās view of the āidealā motorboat taking the best features of the best boats over the last 20 yearsā¦ so I think we can say itās only as good as the prompt you give it and the source data it has to work withā¦ I did ask this one to draw it in the style of Leonardo DeVinciā¦
Sandy
Well-known member
I looked at the early drafts of some university degree courses in Artificial Intelligence and came to the conclusion that it is neither artificial or intelligent. Just a lot of logic gates working very, very quickly.That means you don't know that you are looking at it. Nowadays AI is everywhere. I myself like to ask 'Perplexity' it's opinion on things. Haven't found an answer which I dislike yet
Since then it has morphed into a natural language user interface, in other words it will take a question, trundle off collect some data, it might or might not run some validation on the collected data and present them in a manner that is easy to read. Some will do the same on the spoken word.
In late 2019 I was involved with both Amazon and Google as we were about to move couple of petabytes of dynamic data into their 'clouds' while keeping a service live.
I might suggest that I do know what I am looking at and am still as cynical as I have ever been.
Aja
Well-known member
Aren't fora great? It's soon going to be 2025 and our leading scientist is basing his data on stuff from 2019 and early AI.
God I'm bored.
Sandy
Well-known member
It might be worth reading post #43 in conjunction with post #2.
I'm an engineer not a scientist. I'd find science extremely boring. What is your background?
Frank Holden
Well-known member
I guess one out of four isn't too bad.That's really not surprising given the colossal amount of data being searched and the completely vague search. To use a tool you need to learn how to use it, context makes a massive difference. And your obvious bias just wanted to see a black and white "told you it was rubbish" so stopped there...
"I have a 14m 10 tonne sailing boat in Chile. I want to have it lifted out of the water in a boatyard with hard standing for maintenance. List in detail all possibilities . Give the age of the data as an indicator of accuracy. I am a long distance cruising sailor, add anything else which might be of interest for each result like local spares availability, transport etc for each or anything else which might be of relevance or interest like local tourist activities, trekking etc. Format for easy cut and paste into a forum, do not include any tables but just text"
Quick check on noforeignland & these seem to be go0d enough to be useful.
.....snip
Detail on Club Nautico Reloncavi is OK.
Marina del Sur has zero hard stand , never has had.
Asamar Punta Arenas https://www.asmar.cl/astilleros/magallanes/ has no travelift just a big ship syncrolift. No interest in yachts of 14 metres.
Navic down at Rilan has a big shed that they build quite big boats in together with a ramp. They have 'plans' for 2025.
I would not hold my breath, friend will be down there in a few days and I have asked for an update. They did - pre covid - have 3 moorings but they have sunk and not been replaced. Home - Navik Patagonia
The ones they missed all have travelifts , Antofagasta ( would not recommend), ConCon ( 40 ton, I helped deliver a 50 foot Bene there in 2016 - it was being re-engined - long story ), Algarrobo ( hauled and re-engined my own boat there last year), Valdivia ( can handle cats ), Puerto Natales and Puerto Williams also have quite new lifts although the one in Williams is said to be in a state of disrepair.
I just looked on Noforeignland for the four mentioned by ChatGTP - no mention of two of them and very flaky info re the other two.
Be aware that Navic Patagonia is not Marina Rocas de Rilan. Puerto Marina Rilan en Navily
I think it will be some time before AI replaces LK.
steve yates
Well-known member
And yetā¦. You seen the results from ai on that thread about a boat register. It was mainly a load of bollocks.Except it's incredably useful. Perplexity anyway, I set up a laptop with a new ssd to dual boot today with Linux , be very precise what's wanted then taking photos of the screens on the way if anything didn't make sense, show it the image, "now what do I do" . Bang on replies.
Astounding really.
Showed it a picture of a friend's land rover "what's this" no mention it's even a vehical -" looks like a mark one land rover by the resessed headlights and colour..made between.... " .. Loads of very precise details of how it came to the reply without even being told what the image was.
In 2 seconds blows 99% of the replies on here into touch without the bias and grumpiness.
requiem
Active member
I'm not sure if I've posted the link before, but this is a handy article on how these models work: What Is ChatGPT Doing ā¦ and Why Does It Work? (Warning: it's rather long, but even just the first quarter or so is useful.)
It's certainly quite useful as a "natural language user interface". It works well for that, and also seems to work reasonably well if pointed at something to analyze. As a memory store it falls short, and I think that reflects a mis-understanding of how it works. Neural nets are a fuzzy tool, and thus suited for "fuzzy" tasks. For precision work, other tools are still required. Knowing which to use when remains key.
Always remember, LLMs are not taught concepts or facts but probabilities. One might argue that much of what we humans remember is similarly probabilistic, e.g. if you ask me what Scipio did at Gades you also might not get a correct answer because it hasn't worn sufficient "ruts" in my memory (and it's been a rather long time).
It's certainly quite useful as a "natural language user interface". It works well for that, and also seems to work reasonably well if pointed at something to analyze. As a memory store it falls short, and I think that reflects a mis-understanding of how it works. Neural nets are a fuzzy tool, and thus suited for "fuzzy" tasks. For precision work, other tools are still required. Knowing which to use when remains key.
Always remember, LLMs are not taught concepts or facts but probabilities. One might argue that much of what we humans remember is similarly probabilistic, e.g. if you ask me what Scipio did at Gades you also might not get a correct answer because it hasn't worn sufficient "ruts" in my memory (and it's been a rather long time).
Sandy
Well-known member
Interesting to hear the BBC Radio 4 Today Programme on AI this morning. Not a lot has changed since 2019 or perhaps the media are just catching up? The interesting 'take away' is that the technology is still in the hands of very, very few organisations and even fewer people.
Baggywrinkle
Well-known member
I'm already stuck in the trough of disillusionment .....
lustyd
Well-known member
Using tools is not cheating. Young people have better tools than old people had, but that doesnāt make them cheaters it raises expectations of what they can produce.
Starlink couldnāt have been created with saws, drills and planes no matter how much wood was used. I imagine they used something a good deal better than an abacus for their maths too.
Sandy
Well-known member
That is common practice.
Sandy
Well-known member
If anybody is interested in a practical use of AI then take a look a Prof. Thomas Scott from the University of Bristol work in the use of Spot the robotic dog in Chernobyl, before recent events, and how the team us the Spot to map radioactivity.
The robot is interesting in that you can give it an instruction move from A to B and let it work out the best way from its sensors and is not affected by high doses of radio activity.
Look Inside Chernobyl As Robot-Dog Learns To Sniff Out Radioactivity
The robot is interesting in that you can give it an instruction move from A to B and let it work out the best way from its sensors and is not affected by high doses of radio activity.
Look Inside Chernobyl As Robot-Dog Learns To Sniff Out Radioactivity
Bouba
Well-known member
Wellā¦it is. If you get AI to print you off an essay then it is cheating.
Donāt get me wrong, I hate essay writing (I speak in the present tense because my French teacher still makes me)ā¦.but they are intended for a purposeā¦to show you can research, reason, convey ideas, write intelligently and many other metrics including (in the old days) penmanship.
To shortcut that leaves you either uneducated or unprepared for doing the same during exam conditions
johnalison
Well-known member
I imagine that the world of school and university must be a nightmare fore those trying to uphold ethical standards. One granddaughter was telling me the other day that one piece of her work recently was rejected as having been artificially produced though she asserts that she wrote it herself, which I entirely believe. The irony of course is that the rejection itself will have come from an AI source.
I donāt enjoy AI postings here, but I have no objection if they are accompanied by an appropriate disclaimer.
I donāt enjoy AI postings here, but I have no objection if they are accompanied by an appropriate disclaimer.
lustyd
Well-known member
Well...it isn't. Skills evolve over time. It used to be the case that kids learned the 12 times table. Then we invented calculators and modern kids learn much more complex things instead. They aren't cheating, they're using current tools to their full potential. Using a calculator doesn't leave you uneducated. Quite the opposite, it saves time which can be better spent learning other things.
Most people can't use AI to write a good essay, it's a skill that must be learned. I'd rather kids learn the skill of using the AI tooling so they can go on to bigger, better things than shackle them with the utterly pointless task of learning how to physically write an essay. Many exceptionally bright kids were treated like failures if they were unable to do so in the past (for instance dyslexia, ADD, ADHD, among many others), and those exceptional kids will now feel included and able to contribute as a result of modern tooling. We can now measure kids on actual ability, good ideas, practical thinking rather than ability to present in a way that might not make any sense to them.
ylop
Well-known member
Well certainly if you pass it off as all your own work then it is cheating, just as it would be if you asked a parent to do it for you, copied one your sibling used years before, paid someone on the internet to do it etc.
However, is it cheating if you give ChatGPT the subject of the essay and ask it to come up with a structure? What if you give it your draft essay and ask it to polish up the spelling / grammar? Fit within a word count restriction etc?
Is in fact, the ability to use the tools of 2024 something that should be tested? Knowing what to ask, when to trust it etc is likely a skill that will differentiate the better students. It will make teachers/markers jobs harder - but that doesnāt automatically make it a bad thing. Verbal discussion of the produced work can likely determine in a few minutes how much of what they wrote they really understood.
My āgripesā with āAIā are not that it might be used to cheat or is never useful it certainly can be - but:
- people are treating AI and LLM interchangeably, LLMs are usually very poor at maths, but other AI methods can be exceedingly good at maths, some other AI methods are considerably more energy efficient (if we donāt see the massive data farm needed to turn a stupid question into a stupid answer then itās environmental impact is zero right ?)
- itās a buzzword that every āindustryā seems to be jumping on and investors chucking money at with no idea what they are doing. Crypto was the same - emperorās new clothes.
- users are taking anything AI outputs as though it is more accurate than anything they could have produced by other means. This is just like when people first start using calculators - they switch off all basic sense. LLMs have an extra superpower in tricking people, much like certain politicians, that they can say things in a way that makes it seem convincing, even when it has absolutely no idea what it is talking about.
lustyd
Well-known member
When I was at Uni, the Internet was an identical debate. Eventually academia caught up and clever people are no longer punished for using this amazing resource. When I wrote my dissertation I was marked down for including URLs in the bibliography even though the subject matter was too new for books to have been written.
Give it five years and the old fuddy duddy lecturers will start to catch up with the world. Thankfully, in the real world we just get the job done using the best tools for the job.
lustyd
Well-known member
To be fair, there's no problem with this. GPT-4o added a bunch of skills using other models (including maths) but the LLM fronts them as a human interface (alongside image and speech). The end user doesn't and shouldn't care about the difference, as far as they're concerned AI is a tool that does a thing. It does make sense to differentiate between AI and ML/DS since ML/DS is a human skill while AI is a product. We see this in the segregation of tooling, for instance Microsoft have split AI Foundry (pre-trained things) and ML Studio (for doing data science and training your own models). Customers then use both to make products, but in very different ways.
Bouba
Well-known member
Look ....I am not anti AI....I feel itās unfair that we didnāt have it in my day....all we had was paying another student for essays (which I only did for detention essays)