Comments by "Tony Wilson" (@tonywilson4713) on "AI Deception: How Tech Companies Are Fooling Us" video.

  1. FANTASTIC TO FINALLY SEE PEOPLE CALLING THIS HYPE OUT. I'm an aerospace engineer (by degree) who works in industrial control systems and the staggering amount of nonsense I hear on AI is ridiculous. In the early 1990s I had neural nets and how they actually functioned explained by an electrical engineer who'd worked on them while at University. The origins of neural nets goes back to the 1950s BEFORE the emergence of digital computers. The first generation neural nets were fundamentally analog computing devices set up for complex calculations. The YT channel Veratasium did a good video on what analog computers are, why they were used in the past and why they have potential applications again today. AI (as we know it) is basically digital versions of analog computers and being digital they can be scaled up to immense complexity in a reliable and repeatable way. The most important thing that I see people misunderstanding is that these AI systems DO NOT THINK, but instead they CALCULATE ANSWERS to very complex statistical math. We see this most often when the search box on YT or Google starts giving possible questions as we start typing. Its CALCULATING the most likely search to what you are typing. I BLAME the promoters and media pundits for this misconception.
    212
  2.  @mafiousbj  There's a really good (but less well known) TED talk by an AI programmer from a few years ago. Its about where the training of an AI can go wrong. The talker gave an example of when they tried to get an AI to recognise fish over other marine species. After they taught it instead of recognizing fish it gave bizarre answers. When they asked it to draw a fish it drew the back of 2 human hands. When they went back and checked almost every photo they had showed the AI that it was told "this is a fish" the fish was being held up to the camera by someone. So the most common thing the AI saw in every photo that it was told "this is a fish" had the back of 2 human hands. What that speaker was highlighting was the issue of training AI neural nets with data that can be misinterpreted. There's been a few AI people speak out against the hype and they all have similar stories and all of them rail against the hype. FYI - I work in industrial control systems and have serious concerns that people in management positions will be convinced by the hype-merchants to make some really stupid decisions that we might spend a lot of time undoing.
    10
  3.  @lchpdmq  Please don't argue with an engineer who knows software systems about the difference between calculating and thinking. I don't have that many hours to explain it. First I'd have to explain the differences between digital and analog processing AND THEN the differences between parallel and vector processing AND THEN why quantum computers are going to radically change a couple of areas of technology especially encryption BUT THEN how the new type of encryption called vector encryption will get round that. Bottom line is calculating is NOT thinking although it can at times look like it because 1 area of thinking is actually calculating. Its sort of like cars are a subset of motor vehicles. Cars might be one type of motor vehicle but there are other types as well. Calculating might be one type of thinking but there are many other areas of thinking and most of them we have very little idea of how they work.
    6
  4.  @augustgreig9420  Greta point and it goes to show that almost EVERY INFLUENCER these days is SELECTIVE in what they say what they try and influence. As the old saying goes "follow the money and you'll find a path to the truth."
    6
  5.  @jefflane617  It wont be AI taking the role of the charismatic idiot. It will be AI taking the role of the charismatic idiot's ventriloquist dummy. The problem for most audiences will be recognising which is the dummy and which is the ventriloquist.
    5
  6.  @jefflane617  Agreed - scary, dangerous and NOT good either way its looked at. I see it as very much the like the story of the naked king. Who was imply told he looked fantastic and every one told him he looked fantastic and every one ended up believing he looked fantastic because he wore the finest clothes anyone had ever seen. Then one day a small child simply said "but he has no clothes!" Think about what happened with Theranos. Yeah the PR said it was amazing BUT being able to diagnose EVERYTHING from a single drop of blood was simply science fiction. I'm certain she even got the idea from science fiction (like the Hollywood film "Gattaca"). BUT all that was ignored and she was paraded as the brilliant young gifted entrepreneur who was also a girl taking on Wall Street and the patriarchy. Then one day someone finally said "It doesn't work!" I'm wondering when the same moment will come from AI. My fear it it wont come before doing a staggering amount of damage.
    3
  7.  @AmazingArends  Its NOT mechanical at all, but I get the point. BUT its because it can be the type of system that can play chess as to why it can also take lots of jobs. At any given moment in a chess game the player can chose any of 16 pieces (or less) to move and the available moves are finite. In fact some pieces might not be able to move at all. So the task of choosing the next move becomes definable. As I just described in another comment tasks like image processing for fingerprints and facial recognition are "definable tasks" that by the repetitive nature are things AI is well suited to. Then there's the cases of call centres for things like customer support where there's already stories of job losses.
    2
  8.  @stereo-soulsoundsystem5070  That is a great point and just the other day I heard about how an Indian Call Centre simply sacked over 90% of their people and replaced them with an AI (Chat GPT clone or similar). When you think about how important the call centre industry in India is that's potentially the first step in a major disaster. A lot of those Indian call centre people have spent a lot of money getting degrees in places like Canada and Australia (my country). If that industry suddenly goes away it will also smash the Canadian and Australian university systems because fewer students will come. AND that's before we get into all the other threats to education that AI is promising WITHOUT any proof it can deliver.
    1
  9. ​ @cg.marklemuel  They are a lot closer to taking away millions of simpler jobs like call centers and customer support. Where computer based systems have always worked is on REPETITIVE tasks. What they have managed with these newer systems is to expand that range into things that are not exactly repetitive but are predictable. I mentioned in another comment how I heard that an Indian call center sacked most f its staff and replaced them with an AI. That's the sort of thing that can have dramatic effects with expansive consequences. There are some tasks that AI is particularly good at and there's reasons why. If you think about the text completion in things like search engines. The moment you type the first letter you have suddenly reduced the chances of the what the first word is. Type the second letter and that goes down even further. Complete the first word and start the second word and the number of options reduce dramatically and because of previous searches the rest of the question can be suggested. That's an ideal task for how neural nets can be trained to function and (by extension) also why a lot of customer service work is at risk. Also if you consider things like image analysis tasks like matching finger prints or facial recognition where the task is to compare images or partial images. Its what in engineering can be called a "well defined task." The image can be digitised and key features NUMERICALLY DEFINED into a pattern. Its then a matter of searching for close matches to that numerical pattern. Its a bit like how a person does one of those "Where's Waldo?" photos. Your brain might not just be looking for Waldo but some part of him like his hat. This is why AI is very good at certain tasks or types of tasks. My biggest concern is that its being now PROMOTED as the solution to everything. Think about how many times there's been an announcement of the "next game changing" technology and that's almost the last you ever hear about it. But then we get disasters like Theranos where it gets totally out of hand and people lose out huge.
    1
  10.  @farentimonnaewens4662  Agreed there have been people who warned about all sorts of tech including AI and NOBODY listened. WAY BACK around 1990 Silicon Graphics did a demo of there then latest systems at the college I was doing post-grad. They were one of the first to use RISC based systems and their graphics CPUs (what we now call GPUs) ran at TWICE the clock speed of their main CPUs. it was serious stuff at that time. They were foundered to fill the need for data presentation coming out of the Supercomputers. In those days the CRAY SCs were just giant number crunching beasts. They needed a front end to format the data and a back end to present it AND Silicon Graphics were one of the companies that stepped into that space. They were the first ones to do all those fancy flow visualisations of wings and racing cars. I did aerospace so I got to see that stuff. At that display they showed us they had a 100% created in software short film of a girl walking out of the surf onto a perfect beautiful white sandy beach. Had we NOT been told none of us would have noticed it was a simulation. It really was that crisp. A couple of years later that same company Silicon Graphics became famous for what they did on the film Terminator 2 with the liquid metal robot. Even back then Silicon Graphics and others admitted that in future they would be able to make almost anything and the average person would NOT know the difference. What held it back for so long was how much time (both people and computing) to do even a few seconds. That 10-15 seconds of the girl on the beach (I think) took over a month because it required so much ray tracing. With this new stuff what used to take days or weeks takes minutes and hours. If you recall the film "Running Man" where they map Arnnie Swartz onto old footage with Jesse Ventura for a fight scene. That would have take months back in the 90s now its almost as fast as the film. I have heard Ric Beato (here on YT) talk about the sound equipment they have now that can make anyone into a sensational singer and fix almost any sound track. If somebody hits a wrong note they can fix it in the software. I bet you know more about that stuff than I do.
    1
  11. It also doesn't take much to realise that if he got his wealth from the Dot Com Bubble then it basically means he got out BEFORE the bubble burst. There's NOTHING new about that either. In every bubble there are those few who get out before the bubble bursts and look like geniuses.
    1