Comments by "Joe Qi" (@i6power30) on "Tesla pivots away from Chinese EV 'bloodbath' to autonomous FSD gamble" video.

  1. YEah, but FSD doesn't work. There is no visible timeline it will ever work.
    11
  2.  @sensibility3009  I worked in AI research for 10 past years, I don't see any theoretical frameworks that enables AGI. MIT worked on self driving for several decades, so is Standford. General concensus in academia is that you need AGI for level 4 or 5 full self driving. Tesla's Andre Karpathy leaving in 2022 is an indication that even the industry's best can't find a easy path to full autonomy. Tesla's current FSD lead doesn't have a strong credential or track record. Maybe if the Chinese has an ace up its sleeves, I don't see FSD come to reality for at least 10 years or more.
    7
  3.  @ghrosenb  It will not be solved in a week or two, or a year or two. I promise you that.
    5
  4. FSD is a completely different ball game compared to just scaling to production. This promise might just fall
    4
  5. This kind of story send chills up my spine. You should not be driving if you rely on FSD.
    4
  6.  @ghrosenb  it may seem to work flawlessly for some people because they the the fixed commute route. But there is no guarantee it'll still work just as well if you drive elsewhere or even in the same route a novel situation arises. Humans are more flexible can react to novel never seen before situations.. AI models require millions times more data to train and still stumble sometimes on novel data. That's the danger of assuming it will work because it worked flawlessly in past situations there is no proof that it can react to a novel situation never encountered in its training data
    4
  7.  @NathanBullard-i9k  I didn't say it to never work. I simply said there is no visibility when it will work. Sure if they have something developing in secrecy that's better than anything we already know it might work. I'm just not as confident as you that it will work for sure in near term . To claim that you must know something rest of us don't or a blind faith in Elon musk.
    4
  8.  @kylewagner45  it did improve. Yes it does seem impressive often situations. But it's still not complete.. It's nearly impossible to explain to average people who never worked on AI systems the difference between what seems human like intelligence and true intelligence. It does not progress and improve linearly like everything else. The last 1% could require 10x more effort than the previous 99%
    3
  9.  @coolco1619  Absolutely life takers for some people literally. NHTSA is investigating 460 Tesla FSD / autopilot related accidents, and 13 deaths.
    2
  10.  @hwirtwirt4500  Musk is more an optimistic naive child than a conman. Sadly, many actually think he is genius because he was able to put batteries and electric motors together and attach a tablet onto it.
    1
  11. Just reckless to rely on unproven technology to put lives of others at risk. If you have poor vision you should not be driving at night, FSD still requires interventions.. Your license should be revoked and reported to police
    1
  12.  @thomaswilson2917  them too is not an argument
    1
  13.  @thomaswilson2917  Distracted humans, intoxicated humans, intentionally causing harm humans don't count.
    1
  14.  @doktoronyx6619  The fact that the current neural net models need so much training data is the evidence that our AI model of human brain is flawed. You can train a sub-intelligent human being wit several hundreds of driving data, for maybe a week or two max, he will be able to drive anywhere better than Tesla FSD that had million times more training data, and several years of time, and world's largest super computer. The model is wrong, not the lack of data or compute power.
    1