Slashdot Mirror

← Back to Stories (view on slashdot.org)

Amazon Launches re:MARS Event Focusing on AI, as Second Stage To Invite-only MARS (geekwire.com)

Posted by msmash on from the UP-NEXT dept.

Amazon's annual invitation-only event on machine learning, automation, robotics and space -- known as Mars -- has become a high-tech highlight for insiders, featuring billionaire founder and CEO Jeff Bezos riding a giant robot or walking a robot dog. From a report: Now a wider circle of tech leaders can get in on a spin-off experience called re:MARS, which is due to make its debut June 4-7 at the Aria Resort and Casino in Las Vegas. The event will shine a spotlight on the leading lights and cutting-edge advancements in artificial intelligence and machine learning, Amazon said today in a blog posting.

"We're at the beginning of a golden age of AI. Recent advancements have already led to invention that previously lived in the realm of science fiction -- and we've only scratched the surface of what's possible," Bezos said. "AI is an enabling technology that can improve products and services across all industries. We're excited to create re:MARS, bringing together leaders and builders from diverse areas to share learnings and spark new ideas for future innovation."

19 of 32 comments (clear)

  1. Anyone cares to comment? by ugen · · Score: 1

    I'd like anyone in the know to name 3 things that AI is successfully doing today somewhere in or around our daily lives. Activities that require thinking or reasoning ability and that were once performed by humans, but no longer.

    FWIW I want AI to succeed, but I see a lot of talk and, frankly, close to 0 implementation anywhere I look. So, perhaps, I am looking in all the wrong places?

    1. Re:Anyone cares to comment? by 110010001000 · · Score: 2

      All I know is that when I login to Amazon, it suggests things to me THAT I HAVE ALREADY BOUGHT. How stupid is that? Yeah, I just bought a soldering iron, why would I want to buy another of the exact model? I can understand if it started to suggest things like solder, or an Arduino, or a book on electronics, but it invariably suggests what I already bought. But yeah, AI is right around the corner.

    2. Re:Anyone cares to comment? by ugen · · Score: 1

      Yep, same here. And if I booked a hotel somewhere, I'll see ads for hotels in that same place for a month. Already booked - what's the point? :) It gets funnier. My children are going to middle school now, but since I used to buy baby stuff back in the day, Target (and, occasionally, Amazon) keep offering me more baby items - formula, diapers, you name it. Wouldn't it be obvious that they need to apply a timeline to these types of purchases and offer school supplies after 6 years or so, and then whatever it is teenagers need etc. Instead, it appears to be a static set of keywords, fixed forever.

      Well, so no AI there. But it can't all be marketing hype, can it?

    3. Re:Anyone cares to comment? by 110010001000 · · Score: 1

      Exactly. If AI actually worked, wouldn't you use it for basic stuff in your multi-billion dollar business? Improving the suggestions on Amazon would be very useful and profitable. That is my entire point to AI nutters: if there was AI, why are AI "researchers" using it to play Chess and Go? Why is Siri, Cortana, Alexa so dumb? Could it be that real applications of the technology aren't so easy?

    4. Re:Anyone cares to comment? by Allasard · · Score: 2

      I'd like anyone in the know to name 3 things that AI is successfully doing today somewhere in or around our daily lives. Activities that require thinking or reasoning ability and that were once performed by humans, but no longer.

      NLP (Natural Language Processing): Be it a Voice Assistant(Alexa, Siri) or Voice Prompts while calling customer service.

      Automatic Assistance in Cars: Be it Tesla's automation; or more generic automatic braking assist and lane following. It is still sensor input being evaluated for dangers and a computer taking an action that a human would, but usually faster.

      Advertisements (Deep Learning): Someone I know walked into a store recently, didn't buy anything, and got a physical mailing the next week thanking them for their visit. The data sharing between phone companies, location services, stores and advertisers is a vast network of interconnectedn-ess that is arguably an AI system.

      Computer vision: Camera face identification/tracking and then focusing on that face; Facebook/Snapchat silly face filters; Facebook face identification and name tagging.

      What I learned while getting a degree in Cognitive Science, is that AI is a moving target. What you consider mundane computer assistance today, was the future of AI of a decade ago. It seems like we are never there, but AI is everywhere. Granted, it is domain-specific AI, and not General AI(consciousness), but it is still technically the field of Artificial Intelligence. The goal is to replicate aspects of the human mind, which is advancing rapidly, as seen in the above list.

    5. Re:Anyone cares to comment? by 110010001000 · · Score: 1

      None of that is AI. You are just describing algorithms and computer programs. Everyone acts like this is new stuff. Computer vision and NLP have been around for decades. Processing data is not AI, unless you classify any program as AI. And based on the junk mail I get, and the poorly targeted ads, the AI systems you describe need a lot of work.

    6. Re:Anyone cares to comment? by BlackOverflow · · Score: 1

      Let me tell you how good amazon's AI is - it suggested that we set up a subscription to auto-order a mortar and pestle each month.

    7. Re:Anyone cares to comment? by ugen · · Score: 1

      I am sorry, but none of the things you mentioned (with possible exception of computer vision) have anything to do with AI, not now and not 30 years ago.
      You lost me at "a vast network of interconnectedn-ess that is arguably an AI system." Sorry.

    8. Re:Anyone cares to comment? by 110010001000 · · Score: 1

      Maybe the system is smart enough the know the junk mortar and pestles they sell break regularly. Maybe it is smarter than I thought! I take it all back.

    9. Re:Anyone cares to comment? by SigIO · · Score: 1

      Actually, AI in one form or another has been around for a while. Back in the 80s, my father sold (LISP?) software to American Express, that they used to identify patterns of charges. From these patterns they could discern legitimate transactions from fraudulent ones. I believe it was called the 'Artificial Reasoning Tool' from a company called Inference in California.

      Nowadays, it's somewhat common to get an email or text messages from your credit card provider signalling you of potential fraud. The fact of the matter is, that functionality has been around for 30+ years.

    10. Re:Anyone cares to comment? by 110010001000 · · Score: 1

      That is just statistical processing. It isn't intelligent at all.

    11. Re:Anyone cares to comment? by UnknownSoldier · · Score: 1

      Glorified table lookup is NOT AI no matter how many times you keep spamming it.

    12. Re:Anyone cares to comment? by SigIO · · Score: 1

      I disagree. AMEX had to feed this LISP program real world transactions to train it, and it had to be tailored to individual spending patterns to be useful. Also, perceptrons and weights were implemented in LISP long before they were in Python/Keras.

      Certainly, this application was very focused, and could in no way be considered 'general AI', it was a heck of a lot smarter about detecting fraud than an individual.

    13. Re:Anyone cares to comment? by Kjella · · Score: 2

      I'd like anyone in the know to name 3 things that AI is successfully doing today somewhere in or around our daily lives. Activities that require thinking or reasoning ability and that were once performed by humans, but no longer.

      Once it's happening everybody just says it's an algorithm. For example look at the face/eye tracking in modern cameras, if know I'd struggle a lot of if somebody asked me to write that function. Same with speech recognition, it has other uses too but it's rarely a simple end user product. I know a system that flags certain data for manual processing/review, they of course have hard coded rules too but they also run a more general algorithm that looks for outliers and unusual combinations. I suppose in the first iteration you could call it a simple clustering algorithm, but it's also dynamic with respect to the input as cleared flags means to give more latitude in that direction while confirmed flags tighten it. The bubbles of "acceptable" data flows kinda like in a lava lamp, changing over time.

      I know there's lots of businesses trying to automate processing that way, it's not so much that it handles everything... but it handles all the basic processing that used to be outside "normal" algorithms where you'd constantly run around trying to tweak the business logic and eventually just get lost. I know automated loan approvals go like this, if you're looking a lot like other profitable customers there's literally nobody in the loop anymore. We're still struggling with decision making systems that have to take active, unassisted action like driving a car but they're drawing a lot of insight out of what used to be just static. It still lacks depth but they're mixing basic neural nets with semantic models, memory, training from examples and they're making progress. But it looks like robotics is pretty hard in general.

      --
      Live today, because you never know what tomorrow brings
    14. Re:Anyone cares to comment? by Fnkmaster · · Score: 1

      "AI" generally just means machine learning these days. And all of those are machine learned, statistical model-driven systems, not algorithmic systems. While things like SVM or Naive Bayesian classifiers have been around for years, DNN-based systems that perform well on much harder tasks that require large amounts of data to learn have only existed since 2011 or so.

    15. Re:Anyone cares to comment? by Fnkmaster · · Score: 1

      AI == machine learning. All of these are machine learned systems. AI doesn't mean thinking magic. And yes, we do actually have DNN-based models that do simple reasoning now. Look up the Facebook bAbI challenge, for example, or question answering systems like FlowQA.

  2. Re:Sigh by Spy+Handler · · Score: 1

    Back in the 90's, a large Japanese corporation (Nintendo or Sony, if memory serves) conducted a rather thorough research into this question. The surprising conclusion was that yes, they found evidence of consciousness beyond the brain (psi abilities etc) and it's probably real. But the other conclusion was, they saw no possible way to turn this into a product or profits, so it was quietly dropped.

  3. Re:Sigh by tehcyder · · Score: 1

    Back in the 90's, a large Japanese corporation (Nintendo or Sony, if memory serves) conducted a rather thorough research into this question. The surprising conclusion was that yes, they found evidence of consciousness beyond the brain (psi abilities etc) and it's probably real. But the other conclusion was, they saw no possible way to turn this into a product or profits, so it was quietly dropped.

    If you can't think of a way of turning actual psi abilities into profit, you're really not trying very hard.

    --
    To have a right to do a thing is not at all the same as to be right in doing it
  4. Re:Sigh by Spy+Handler · · Score: 1

    They didn't develop psi abilities, they just saw enough abnormalities to suspect that it exists.

    Seeing evidence of, and suspecting that the sun is powered by nuclear fusion does not mean you can suddenly start building fusion reactors.

    Suppose you ran an experiment where 10 different objects were hidden in a box and test subjects had to guess which one it was. You would expect 10% correct results over the long run. Now take another group and have them meditate or pray to cosmic forces and strongly visualize the box for 2 hours before doing the experiment, and you got 18% correct results.

    Now suppose your trial size was very large and your test protocols were rigorous and scientifically beyond reproach. That 18% becomes statistically very meaningful, it is strong evidence of psi abilities. But how do you profit from this? How do you create a product bases on this information? You can't.

    I just googled it, it was Sony's project (not Nintendo) called ESPER, and they actually quietly buried this project after it was concluded. They most definitely did not shout it from rooftops. However the existence of this project leaked out to the media at some point and a Sony spokeman was eventually coaxed into giving this statement:

    Sony spokesman Masanobu Sakaguchi, to the South China Morning Post after the story about the company's research broke: "We found out experimentally that yes, ESP exists, but that any practical application of this knowledge is not likely in the foreseeable future."