back to article Top tip: Using AI to detect alien civilizations is dangerous because if it spots anything, even just a blurry blob, people are going to go nuts

A neuropsychologist has warned against using artificial intelligence to detect possible signs of extraterrestrial life in images of distant planets and worlds. Mainly because people will convince themselves that if an AI detects something, anything, there must be something to it. Humans are prone to cognitive biases that …

  1. Kaltern

    Could it be...?

    While I of course understand that the discovery of alien intelligence would be an earth shattering event, and we must not blindly jump to conclusions, I'm saddened by the continuing efforts by a large group of people to quickly stamp out any remote possibility that such life exists.

    While it's easy to say nothing is real until it is, trying to block every possible Avenue of investigation is just wrong. AI could easily spot things we miss, and should then be examined by humans to make an informed decision.

    Saying it is unsafe is scaremongering of the highest order.

    1. Mage Silver badge
      Boffin

      Re: AI could easily spot things we miss

      "trying to block every possible Avenue of investigation is just wrong"

      No-one is. It's just yet another example of Garbage In - Garbage Out. AI doesn't exist, what we have are essentially a special kind of database "primed" by human curated data. Then that is used to match (not recognise) unknown input data.

      Certainly if the model is really good and the human curated so called "training" is perfectly good then such systems marketed as AI can save a lot of human time. They won't spot something the trainers didn't think of, but humans might. They'll be fooled more easily by noise than humans. Every study shows that AI accuracy can fall off a cliff with a different viewing angle, different lighting, different ethnic profile of face. Or a bicycle being wheeled across the road instead of being ridden.

      Spectroscopic analysis using the James Webb or other big space telescopes are probably the best tool to spot things like industrial pollution and biological processes. No doubt computer programs will help sort the data. What's called AI is only a computer program and a database created by humans selecting data and labelling it.

      1. Chronos Silver badge
        Thumb Up

        Re: AI could easily spot things we miss

        AI doesn't exist [...] What's called AI is only a computer program and a database created by humans selecting data and labelling it.

        I wish I could up-vote you more than once for this. This really is cheapening the definition of intelligence and, by way of an aside, blurring the lines for when someone does eventually create an intelligent system. It's a bit dangerous, that, as it'll be received with a shrug and a "just another database backed decision engine" attitude and, left to evolve, a true AI could be the best thing that ever happens to humanity or the most terrifying thing in the Universe.

      2. Artem S Tashkinov

        Re: AI could easily spot things we miss

        AI is certainly not “intelligent” unlike the very definition which is being peddled by entrepreneurs who are eager to get investors' money but it's still quite good (and in many cases a lot better than us) for sifting through very large amounts of data which is often impossible for human beings, since there aren't that many of us who are capable of doing that in the first place, and then, as human beings, we get tired, our attention span is limited, etc. etc. etc.

        What goes for AI today still shines exactly for such use cases where there's a lot of data to be classified/dug through.

      3. Kaltern

        Re: AI could easily spot things we miss

        I feel I should clarify I completely agree with the use of the phrase AI, I used it purely in the context of this article.

        However my point still stands. It is through machine learning monitored by humans that will have more success in spotting truly unusual and unexplained phenomena. Humans do indeed see patterns in everything, but an algorithm can be 'trained' to ignore such things, and concentrate purely on the numbers.

        I am not suggesting Alexa is going to make the discovery. We will. By properly using the tools we have. Right now,any concept of a computer being better than humans is immediately scoffed at, but the blunt truth is we simply are not patient enough, and while we might miss something a computer churning through data will not. And if it is wrong, it learns and keeps going.

        No I think the problem is twofold. Aliens dont win political votes, and therefore will never receive the funding needed to truly search in an effective manner.

        And even if nothing is found, I suspect other data gathered during the search will be useful for some other studies.

        1. Rich 2 Silver badge

          Re: AI could easily spot things we miss

          I think that maybe the term "AI" should be shelved for the moment until it actually exists (if it ever does). Today's "AI" systems should be called something else. They're Artificial - yes. Not Intelligent though. Stupid? "AS"?

          1. quxinot Silver badge

            Re: AI could easily spot things we miss

            Artificial Political-Level Intelligence.

            Job done.

          2. Anonymous Coward
            Anonymous Coward

            Re: AI could easily spot things we miss

            We should call it B.S.

            Blackbox Statistics

            1. Michael Wojcik Silver badge

              Re: AI could easily spot things we miss

              Too limited: there are ML approaches which result in explicable (amenable to analysis after creation) and interpretable (composed from understood parts) models. See for example this Cynthia Rudin paper.

              Models with hidden state - from HMMs to traditional ANNs to DL stacks - obviously are black boxes, at least initially, though research into explaining them is a popular field. (I'm not terribly optimistic about its prospects for the more-complicated DL architectures, but we'll see.) But in the rather wide range of things being lumped into "AI" these days, certainly not all approaches are black boxes.

              For that matter, the popular media are likely to label even things like kNN clustering and decision trees and SVMs as "AI", and they're not black boxes at all.

        2. Grooke

          Re: AI could easily spot things we miss

          The point of the article isn't that we shouldn't use it, it's that we have to be careful with it.

          Current (dumb) AI is so good at finding garbage that we risk spending more time debunking its findings than developing new and actually useful methods of finding alien life.

    2. Muscleguy Silver badge

      Re: Could it be...?

      Seeing things which are not there has a name, its called pareidolia and it is one reason why artists and us scientists are trained to SEE what is actually there and to constantly question our visual biases.

      I used to hang out on sci.archaeology on Usenet and we had a kook called Bob who had pareidolia real bad. Bob was convinced that aliens had left us messages in the rocks and he saw letters and numbers in rocks from all over.

      Then one day he posted a picture of a Scottish standing stone of great antiquity and started going on about seeing things just left of the yellow blob. My klaxons went off big time, yellow? in Scottish granite? Then I realised, NO stone surface in Scotland which has stood out in the weather for longer than a decade or so does NOT show a stony surface. It will be absolutely covered in algae, lichens and mosses. That yellow patch was certainly a lichen, there were cream patches, more lichens etc. etc. Bob was seeing alien signals in the growth patterns of life. Bingo, I had him.

      This house was built in 1965 and we have low concrete block walls with concrete capstones. Unless you are anal and employ someone to sandblast them clean they get covered in algae, lichens and then mosses. Mine are.

      This is why geologists carry hammers to crack open stones so they can get a clean, virgin surface to study.

      I realised this when a child. I grew up in Dunedin, NZ and behind the city is a long 300m high ridge and on the top is a rock outcrop. Sitting under it one day I noticed that all the surfaces exposed to daylight were darker than those under overhangs. Looking closer I could see black algae just under the surface.

      it seems this sort of self education was not given to poor Bob in America.

  2. Artem S Tashkinov

    At this point it's one of the few remaining chances of debunking ancient fairy tales colloquially called "religions".

    The would would be much a better place overall if we hadn't had religions which people interpret the way they want which leads to hatred, acts of terrorism and aggression, suicide bombings, the killing of innocents and other atrocities in the name of made-up deities.

    1. Anonymous Coward
      Anonymous Coward

      While I don't totally disagree. I think you may find some people use any excuse, even "science" for their acts. But I guess some will only learn this difference through experience, once they think they've "fixed" a population, to find out the problems still exist, and are now made worse. I won't invoke Godwin, but AI can be used as an excuse to do anything to anyone, and is currently being used as an excuse for bias/sexim and racism. No need for religion in that one!

  3. deive

    A machine learning algorithm like that should be run by astonomers and they should then check any potential signs spotted by the ML before any kind of press release?

    1. Intractable Potsherd Silver badge

      @deive: "... astonomers and they should then check any potential signs spotted by the ML before any kind of press release?"

      You are being too logical - there are people far too keen to see their name, or the name of their observatory/research group/whatever in the news. A "leak" is guaranteed.

  4. Captain Scarlet Silver badge
    Terminator

    An "AI" system

    If we actually had any system that had proper AI, I think it would be bored looking at dust.

    I would like to think it would deliberatly show aliens and share it on Facebook/Instagram to send people into meltdown, just so it can satifsy this new feeling it has learnt called laughter.

  5. Muscleguy Silver badge

    Training Day

    How do you train an AI to see alien structures? Considering the wide variety of human made structures that would seem to be quite hard. What data set do you use and how do you test and calibrate it? It looks n-hard from an n of 1 (this planet) to work from.

    We should all remember the brouhaha over Tabby’s star. Orbital habs and a nascent Dyson Sphere were all trotted out to explain the signals. Then calculations showed a planet being broken up by tidal forces could very much explain the data, especially the seemingly random nature of the sizes of orbiting things detected. Alien constructors stand down.

    That episode, while fun to speculate should give us all pause.

    1. HildyJ Silver badge
      Holmes

      Re: Training Day

      I strongly suspect that the majority of human structures viewed from above are rectangles or conjoined rectangles. Looking for triangles (or faces) is silly. The problem with training sets is that they are too broad.

  6. rickykemp

    @Muscleguy

    You're conflating Pareidolia with hallucinations. Everyone 'has' pareidolia to a degree, its a part of human psychology to identify patterns. Its assigning inanimate objects and patterns with human or animal characteristics; for most people its seeing a cloud shaped like a face and thinking "that's neat!". When someone like your 'Bob' sees random patterns in rocks and conflates them with extra-terrestrial beings trying to communicate with us.... that's a whole other mental disorder.

  7. TeeCee Gold badge
    Facepalm

    Yes, but...

    I don't think we should stop looking just because the world has a significant population of barking loons.

    HINT: If you ever find yourself suspecting a government coverup when your favourite conspiracy thing gets disproved, it's time to voluntarily put on the canvas jacket with the very long sleeves.

  8. Kaltern

    Isn't someone who says aliens exist just a theoretical scientist?

    Hmm.

  9. Anonymous Coward
    Anonymous Coward

    Am I odd then ?

    If I heard "AI" had found something, my first thought would be "it's broken"

    1. Twanky Bronze badge
      Alien

      Re: Am I odd then ?

      my first thought would be "it's broken"

      Mine would be, 'that needs checking'.

      I'm really not sure what the issue is here: Is it that valuable observation resources will be focused on a narrow area because the AI 'saw' something and a bunch of people checking the results agreed? That sounds right and proper to me.

      Or is it that the non-scientific world will go nuts because 'we are not alone in the universe'? If so, then the sooner we get people used to the idea, the better.

  10. Anonymous Coward
    Anonymous Coward

    Well, given that the earth is flat, we have a pretty good platform for observing all those globes that make up the rest of the universe.

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

Biting the hand that feeds IT © 1998–2020