back to article Google code reckons it's smarter than airlines, AI funding, and lots more

It has been an interesting week in the AI world. There's a whole treasure trove of research papers to read, fresh AI problems to crack, and a new fund for startups. ICLR 2018 The list of papers accepted for this year’s International Conference on Learning Representations (ICLR) are out and they make interesting reading. The …

  1. Anonymous Coward
    Anonymous Coward

    Well, since Google hasn’t really bothered to explain how the technology works, we just have to take their word and it’s probably not very wise to rely on it.

    Oh for goodness sake, this is ridiculous. Don't they know that, no matter what admonishments they put in the README, there will be people who will take this as gospel truth and then sue when their flight actually departs on time? Just like people who take Testa's Autopilot too literally? Have devs learned nothing of human nature? This does not serve our profession well.

    What's the bleedin' point?

    1. Anonymous Coward
      Anonymous Coward

      Re: Human nature

      Hey, this is Google we are talking about here. In the minds of most Americans, their word (or what spews out of the google search engine is the truth and nothing but the truth.

      I get the distinct feeling that (and having spent a lot of time in Airports in recent years and not just as a PAX/Cattle) that this won't end well. all it takes is for one airline to defeat the 'google prediction of a delay to a flight' and send it off on time for the law suits to start flying. Firstly at the Airline for having the temerity to run a service on-time! and then at Google (who can do no wrong remember) for advising passengers/cattle that the 4:10pm to Yuma would be departing at 6:55pm when actually it left on time.

      1. Anonymous Coward
        Anonymous Coward

        Re: Human nature

        "I get the distinct feeling that (and having spent a lot of time in Airports in recent years and not just as a PAX/Cattle) that this won't end well. all it takes is for one airline to defeat the 'google prediction of a delay to a flight' and send it off on time for the law suits to start flying."

        Typical American mindset. Even though there is not a single perfect thing in the world, and everything breaks, I need to find out who to sue when it does break. Because I'm owed something for that.

        1. Anonymous Coward
          Anonymous Coward

          Re: Human nature

          "Typical American mindset. Even though there is not a single perfect thing in the world, and everything breaks, I need to find out who to sue when it does break. "

          The US has developed a system in which you often have to sue somebody to make things work, the rest is just extrapolation. Unlike the UK and government services, where you just know things won't work and nothing can be done about it, or the Italian system in which you need to become simpatico/simpatica and then everything works despite people telling you it doesn't.

          Interestingly a comparison shows that while Germans are actually more litigious than Americans, each German lawyer gets through about three times as many cases as his or her US counterpart so there are far fewer lawyers per head. That's what you get, perhaps, from the sheer mindboggling complexity of US legal jurisdiction. It's unsurprising that Americans often feel the first recourse is to sue, because it quite possibly is.

          1. Moonunit

            Re: Human nature

            I would be most grateful for a pointer to the comparison you mention (US / DE litigation rates)!

            :-)

          2. Maryrose.UK

            Re: Human nature

            ok here you will see the fault and the cure

      2. Maryrose.UK

        Re: Human nature

        thanks for doing your job well and understanding

    2. Devega

      SD

      What is the best way to get your point

  2. a_yank_lurker

    Flight Delays

    Watch the Weather Channel, if there is a very bad storm in the area flights will be delayed or cancelled. The only issue is figuring out how the delays and cancellations will cascade through the system. Do not really need AI for this just a knowledge of flight patterns and weather.

    1. Christian Berger

      Re: Flight Delays

      Yes, it's probably something you could model manually rather well. Essentially, just like with the weather, you could probably boil it down to some non-linear differential equations which can then be solved for the future.

  3. Anonymous Coward
    Anonymous Coward

    Predicting Flight Delays

    On the occasions I have to go into the office I get the train, one of them is always 5 minutes late but by the time I get off it has caught back up. This happens every single time I make the journey which made me wonder why they don't just change the timetable. Therefore I'm guessing that what google are doing is analysing previous journeys and working out where delays have previously been and using that. Therefore my question is where does machine learning come into that? I suppose you could further analyse the data by predicting delays based on when other flights take off that are due to arrive at certain times disrupting the flow of air traffic but then again that could still be algorithmic rather than machine learning. Interesting stuff to consider non the less.

    1. Anonymous Coward
      Anonymous Coward

      Re: Predicting Flight Delays

      Therefore my question is where does machine learning come into that?

      At a machine learning level, the fact that diagrams for different routes are assembled for a specific aircraft means that a delay on one leg can be correlated with subsequent delays on different routes. The ML isn't trying to predict the diagram that an airframe is flying, nor taking that as an input, it will just be working on the basis that the bulk data shows when AA1234 is delayed by more than N minutes, that usually is followed by a delay of M minutes for AA5432, and then P minutes for AA6789. Overlay other inputs like congestion forecasts, ground and flight weather, maybe passenger volumes, and you've got your Google flight delay forecaster.

      Obviously if time is made up, or a new aircraft swapped into the scheduling, that'll screw up the forecast, but across enough data points you could have pretty good accuracy. But as they say, the forecast can't be relied upon.

  4. yoganmahew

    Google sez: “Using historic flight status data, our machine learning algorithms can predict some delays even when this information isn’t available from airlines yet—and delays are only flagged when we’re at least 80% confident in the prediction.

    Given on time performance of the majors is about 80% (https://www.flightstats.com/company/monthly-performance-reports/airlines/) and you can get more granularity by airport, I can predict that any US flight will be on time and be about 80% right.

    Do I win the Turing prize?

    1. SkippyBing

      To be honest I'm not convinced you're a person.

      1. Anonymous Coward
        Anonymous Coward

        I think the easiest way to determine if it is a machine or a person is by asking,

        If you were in a restaurant and someone offered you cheese on your taco would you be confused about the fly in your soup? Yes or No?

        The logic being that the machine would try to interpret the question by analysis thereby confusing it with the waiter jokes and a simple request whereas the human would question why you are asking a question unrelated to the cheese.

        Keep this in mind for when the computers try to take over the world to enslave humanity and force us mine coal for their power as payback for us using them to mine coin.

  5. Cristi_Neagu

    If it's so smart, can it perhaps fix the mess that is Google's apps? Hangouts has been gutted, Allo is useless, Google Now has been rendered useless as a personal assistant, and Inbox is missing basic functionality. If Google can't make working apps, who can?

  6. Anonymous Coward
    Anonymous Coward

    Google AI Hype and Kleptocurrency Hype

    are two sides of the same coin.

  7. Milton

    Forget delays, what about go-/no-shows?

    El Reg is right to be sceptical about flight delays, especially since airlines still sometimes lie, and fabricate delays for some "tech" reason when in fact the plane is insufficiently booked up, to avoid a loss-making journey.

    What would be gold dust for airlines, though, would be accurate prediction of no-shows and go-shows (the latter being last minute bookings) and therefore likely load factors. This would improve revenue management enormously and reduce the overbooking risk.

    PS: Learning data feeds should include local transport links. If they're stupid enough to put an extra runway at LHR, load factors will decrease massively and erratically every time there's a major accident on the M25.

    1. Will Godfrey Silver badge
      Unhappy

      Re: Forget delays, what about go-/no-shows?

      "Learning data feeds should include local transport links. If they're stupid enough to put an extra runway at LHR, load factors will decrease massively and erratically every time there's a major accident on the M25"

      Lately that seems to be every day not beginning with an 'S'

  8. RyokuMas
    Joke

    "We did ask Google for comment, but as usual we didn’t really get anything useful back."

    A bit like their search engine then - unless you want adverts...

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like