Oh no, startup Massive Analytic unleashes 'artificial precognition' When Britons do tech startups they don't hold back. London-based Massive Analytic is an artificial intelligence startup that has created Oscar AP, a product they describe as 'artificial precognition'. If name Massive Analytic wasn't just this side of disturbing, the unnecessarily creepy "artificial precognition" marketing …
... then I quickly forsee a lot of El Reg and other online tech-savvy users doing their damndest to poison the database with spurious information, not to mention a massive new market for cookie blockers, script blockers, ad blockers and anything else that will ruin their business plan ASAP...
Signed - Ethel A Aardvark, aged 63, from Tunbridge Wells...
Expect governments to be all over this. Your local council would like nothing better than to find more efficient ways to fine you for littering, jaywalking, putting out too many bags of trash and anything else they can dream up.
Um, it does still require there to be someone sufficiently cognisant to successfully formulate a question 'though. Bugger all chance of that around my local council.
Really? I think most of the tools I own have had other instances that, at one point or another, were turned against someone. Whether it's a sharpened screwdriver used as a shiv or a hammer used to build a gallows, certainly few hand tools have never been turned to dubious ends. And the same can be said of cars, clocks, cooktops - take your pick.
I can't think of a tool that can only be used for good purposes. (And "good" as defined by whom?)
If AI ever progresses to the point that this is possible
Where what is possible? I saw only the vaguest of claims in the article. (I admit I skipped the marketing video.) Maybe when the patents are filed we'll know more; they may also have whitepapers or something on their site.
But what I read in the article seemed like straightforward evolution of the "analytics for non-specialists" trend that's been developing for a while now. These folks may well have done a good job gluing pieces together (particularly in the "both structured and unstructured" area; there's been a ton of work there, but that doesn't mean it's easy or common yet), and they may well have some innovative techniques.
In the broad sense, though, the basic toolchain of digest data, build a model, make predictions is well understood and plenty of people are doing it already. For business, the "make predictions" part has to give results that are sufficiently better than what you're currently using to forecast in order to be worth doing; but that's generally not a very high bar.
In itself, "artificial precognition" doesn't mean squat.
"Expect governments to be all over this. Your local council would like nothing better than to find more efficient ways to fine you for littering, jaywalking,"
Jaywalking? Not in the UK, we still presume (albeit erroneously in some cases, but generally works OK) that people are capable of crossing roads without being at a crossing / when a crossing has "walk" sign on red. With the pedestrian unfriendly road layouts, few crossings and long wait for "green man" at crossings, without a bit of "jaywalking" you would be taking hours to do a journey of any distance in most UK cities
Although probably about 2 years ago now, I had a preview of Massive Analytics' solution. At the time, it looked a bit hashed together, and wasn't a particularly interesting proposition. I don't know if things have changed since, but this looks like a poorly disguised press release, and very unlike the Register to punt company trade marks around unquestioningly.
Disclaimer: I'm not a competitor, just someone who wasn't particularly impressed with marketing hype.
all in the hope of getting more investment money leading to an IPO at which time several bods may well depart leaving the company to flounder.
I'm not pointing my finger at this company specifically but far too often we hear of things like this and a few people ride off into the sunset with a shed load of cash and everything dies a death.
I'm old enought to remember the massive hype around a thing called 'The Last One'. Full page adverts in the Computer rags of the time. Hyped as the last computer programme you would even need to buy. All it was a program that generated Basic!
Then we had the .com bubble. Now? the Big Data bubble?
Just to be safe, I'm posting as AC.
some final thoughts
After all Big Data is everywhere ain't it?
It has gotta feed its appetite somehow.
How much data does this outfit have on you?
Ever done something a bit dodgy? Well who knows this sort of company may have a record of it.
Ashley Maddison? Pah small fry.
etc.
"Just to be safe, I'm posting as AC."
Now you see, if you had this Oscar AP thingy at your disposal (or either Samaritan or "The Machine" from TV's Person of Interest!), you wouldn't have needed to hedge like that. You could have reliably predicted not only how many up and down votes your comment would generate, but who would make them and even when! (And then modified the content of your post accordingly.) Hell, you even would have foreseen my response!
Everybody knows computers can do this sort of thing now, right?
.. my experience suggests that someone has been rather over zealous with the claims here.
The issue here is that most data collection simply doesn't have enough information to draw deep conclusions about end users, beyond simple 'people who like X also like Y' relationships. The problem is that most companies' views of users are restricted to browser sessions and occasional logins, and most interactions are of the form 'looked at X, bought Y'. The restriction here is that you don't know who is actually behind the keyboard at any given time, and inferring the reasons for their choices has to be based on extremely limited information.
Hence I visit Amazon regularly, buy items for niece and nephews' birthdays and occasional needs for my own wife and kids. In a recent list of 'recommended for you' I had a crochet kit and a hand axe besides each other - both utterly irrelevant and not reflecting the actual purpose of my visit that day or even the following year. Worse still, that's for a site that I visit (depressingly) regularly. Most sites suffer from customer loyalty that barely registers on the chart, meaning predictions have to be based on little more than the time of day and the location you logged in from.
Now undoubtedly you can improve the accuracy and timeliness of recommendations (the base level being random guesses from your marketing department), but the vision of precondition and overthrowing governments is far from the truth.
"In a recent list of 'recommended for you' I had a crochet kit and a hand axe besides each other - both utterly irrelevant and not reflecting the actual purpose of my visit that day or even the following year."
I purchased a PA amp for manufacturing facility a while back. The next three months I kept getting ads for (basically) the same product. This is not a "consumable" (I suppose that depends upon it's quality). My guess is that marketing, the spawn of consumerism, looking to justify itself, will show (with tweaked data) that I bought this because they advertised, not the other way around. They'll do anything to look good when it has been... just... failure after failure after nobody believes they're worth the investment anymore...
Ah the awful stench of Marketing Bullshit. This time, the bullshit was written by Marketing, to sell something to.... Marketing. "Give us your big data and we can infer all kinds of crazy sh*t from it that you can neither deny nor confirm". Consumerism and Capitalism all rolled into one. It would make me run a mile. Or ten.
By the way, I think you'll find the term is 'Eke out" rather than "Eek out", however the phrase "Eek! ......Out!!!!" is very apt in this particular case.
Just be prepared and you should be ok.
1) Wear a disguise so that the facial recog cameras can't ID you. go too far and you might get ID's as a terrorist though.
2) Turn any mobile device you have either Off or into Flight mode. Then the beacons can't track you and beam targetted Adverts at you.
3) Pay for everything with Cash then the Credit card computers can't know what you bought.
Otherwise,
Have a nice Day! :)
The thing about software like this is that, in general, it doesn't work very well on real world data. Real data is complicated, untidy and it is very hard to turn it into anything informative even if you know how to interpret language and tie together the data in this excel spreadsheet, that long-abandoned CRM database, another database which someone made as a spreadsheet because they don't know what databases do, and that huge archive of word documents and scanned PDFs.
Even working with curated and well designed data, big data inference tools are as likely to be stumped or to raise spurious correlations which require a lot of statistical knowledge to verify or disprove.
I anticipate a long chain of disappointed customers and a start-up selling software that never quite does what their sales team claim until they run out of rubes to sell to and quietly vanish beneath the waves.
I disagree actually, I would say they need to be better than not using them. However I doubt that any of the data tools around are better than nothing at all, let alone better than a well designed marketing strategy based by regular market research. At some point they will be, but then they'll end up in the market researchers' toolbox anyway.
To Tesco Clubcard then.
They find your patterns and offer you vouchers to match those patterns.
Fortunately for them they have a closed data set which makes this easier but can still be skewed if like me you purchase stuff for an office and buy lots of coffee, fruit and biscuits from them which masks a lot of my normal shopping habits.
If someone walks into a shopping mall and the system can see this its pretty obvious why you are visiting so spamming ads at specific geolocations is probably about as far as future predictions get, after all, you dont go into a shoe shop to buy toothpaste!
We already have the Avatar:
Known for: "Making erroneous predictions, predicting the future on TV, radio, books, records and in newspapers and as a narrator and actor in the films of Ed Wood"
Just a couple of hours before I read this story, WGN commenced rebroadcasting Season One of one of my favorite TV shows, CBS's hit show Person of Interest. I plan to start watching the first three episodes in about an hour!
(Sadly, I was unaware of the show's existence when its first season was broadcast. I only caught on — and got hooked — during Season Two. I've been waiting several years to see Season One!
The ways in which the show exaggerates the capabilities of computers and AI are obvious to me, but even the things that I consider unrealistic are done in an intelligent and convincing way, grounded in reality. Suspending my disbelief is easy!
Especially since the heroes on the show are all acutely aware of and on guard against the dangers of the technology and the potential for its abuse. While the villains — and especially the chief villain, played so masterfully by the show's creator's uncle! — just can't wait to welcome our new cybernetic overlord. (Just as on Highlander, there can be only one!) I have long conjectured that this show has to be the favorite show of everyone who works at the EFF. If they don't hold weekly viewing parties, something is wrong!
Obviously the point of this entire comment will be utterly lost on anyone who is not familiar with the show.)
So: Just a coincidence? Synchronicity? Kismet?
The expression "artificial Intelligence" always reminds me of a Goon Show episode:
<Neddy Seagoon is rescued from the water once again>
Peter Sellers: He is almost drownded - give him artificial respiration!
Spike Milligan: Artificial respiration? The man is dying, dammit! He needs the real thing!
"Artificial respiration? The man is dying, dammit! He needs the real thing!"
But in the case of "artificial precognition", there almost certainly is no "real thing"!
(Indeed, it seems far more likely and plausible that there might be meaningful probabilistic predictions of people's future decisions and behavior gleaned by machines from high-speed mining and analysis of masses of data, and the detection of hidden patterns and correlations, than that there might be any sort of actual precognition of future events by humans using psychic powers. On this particular question I agree entirely with Patrick Jane of the TV show The Mentalist: "There's no such thing as psychics" (1:56), but there are people who can quickly make good educated guesses that sensibly work the odds.)
From the Paythru quote: 'Oscar AP’s approach to analytics would allow us to target our customers better and quicker'
Note use of the word 'would' - so this hasn't actually been deployed in the wild yet? Also, the Paythru mouthpiece states that they are development partners with Massive Analytics so presumably have a vested interest in garnering support for the tech in order to get it finished.
I'm with all the commentards who point out the limitations of real world data. I suspect that the machine-inference success rate will not be high enough to be a practical precog system for anything other than improved marketing in some subset of cases where the data is particularly good.
It's a marketing hype piece. Not to say it won't happen, just that the paranoia scare piece that forms the major slant of this article is wrong (for the moment at least)
>If the technology is even a fraction as good as advertised, Massive Analytic is very soon going to have more business than they know what to do with.
That's alright, if the technology is even a fraction as good as advertised, Massive Analytic will be able to predict which customers are worth having.
It sounds like interesting tech, and much of this is possible in principle,* but the story is awfully long on claims and short on substantiation. Indeed, it doesn't even attempt to provide any explanation of how Oscar AP is supposed to differ from the countless other attempts at predictive analytics or artificial intelligence! It just gives us a new buzzword ("artificial precognition").
.
* To be precise: The detection of otherwise-hidden correlations certainly is possible, and that information can often be quite useful and illuminating. Predictions of future behavior, on the other hand — whether of complex systems or of individual human beings — are another thing altogether. While some types of limited, probabilistic predictions might well grow out of the patterns and correlations that are detected, it is highly doubtful that — for instance — Oscar AP will ever enable "Elected officials . . . to know who dissidents are . . . before the dissidents themselves know they are going to speak up." That sort of claim is as inherently problematic (if not ridiculous!) here as it was in the story Minority Report. I predict (irony intended) many lawsuits if Oscar AP is used to "spot personnel issues and . . . predict failures in . . . people."
Legend has it that in a graduate economics class at The University of Chicago, after Milton Friedman had put on the blackboard a single statistical equation intended to describe the entire economy, a student called out "What about free will?" Professor Friedman reportedly turned back to the board, added an error term (plus-or-minus epsilon, "±ε") to the end of the right-hand side of the equation, and then turned back to the student and said "There, that's free will!"
(Anyone who finds this story at odds with his perception of Milton Friedman needs to realize that Friedman was just as committed to positivist methodology as he was to libertarian ideology.)
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
