You Look Like a Thing and I Love You: A quirky investigation into why AI does not always work

Everyday AI has the approximate intelligence of an earthworm, according to Janelle Shane, a research scientist at the University of Colorado but better known as an AI blogger. Since AI is both complicated and massively hyped, and therefore widely misunderstood, her new book is a useful corrective. You Look Like a Thing and I …

[John Smith 19]

The trouble is Artificial neural networks are very attractive.

After humans are neural networks and we're intelligent, right?

But human NN can evolve

ANN's don't. Can they identify something they've already seen before? Probably (because that's what they work on, probability)

Can they create something they have never seen before? Probably

Is it going to be useful/attractive/pleasant/safe? F**k knows.

My instinct is that human NN are actually the the host for evolved microcoded processes that are too dynamic for current imaging to identify. IOW ANN's are just too primitive (and because they cannot evolve will always be too primitive) to ever be anything but occasionally useful. Their danger is they look much more impressive (in carefully controlled scenarios) than they actually are.

So just clever enough to be deployed in the real world and hence very dangerous indeed.

[amanfromMars 1]

Because its complexity is simpler to harness?

Are you struggling to make machines more like humans when we should could be making humans more like machines?

[amanfromMars 1]

Who needs to get out more :-) or is that just for Trivial Pursuits and Spooky Actions at a Distance?

Posts by John Smith 19 15628 posts • joined 10 Jun 2009

Posts by amanfromMars 1 5797 posts • joined 10 Jun 2009

:-) Is El Reg addictive and as a gateway drug to deeper and darker, higher and brighter levels of existence? Enquiring and/or addled minds may wish to know for the comforts that reward one with glorious confirmation. :-)

[amanfromMars 1]

Re: Who needs to get out more for Trivial Pursuits and Spooky Actions at a Distance?

I'm now a'wondering if one of the main criteria for the gold badging of a virtually anonymous and presumably real poster, is simple post quantity or a more complex quality ..... with an amalgam of both yet another fine root and route to take ....... for shortcuts and fast tracks.

[John Smith 19]

@amanfromMars 1

No, still no idea what it's talking about.

Perhaps you should check the house rules for badges?

[amanfromMars 1]

Let there be Light

No, still no idea what it's talking about.

Perhaps you should check the house rules for badges? .... John Smith 19

It would appear to be very much a case of what is said, understood and liked enough by both El Reg and the masses being bothered to register an upvote, John Smith 19, for El Reg qualifying thresholds for badges surely confirm it? So mere prodigious quantity bears no responsibility.

Bronze ... More than one year members and more than 100 posts in the last 12 months.

Silver ..... Silver badge holders meet bronze requirements and have more than 2000 upvotes.

Gold ......This discretionary badge is awarded by Reg staff to commentards who have been very helpful - to us, through news tips and beta testing, for example - and to their fellow readers, through their posts. ....... https://www.theregister.co.uk/2012/03/23/el_reg_forums_faq/

However, discretionary awards are always liable to be thought easily subjected to bias rather than solely being well considered as a truly worthy reward to certain egos/ids.

:-) And one must never forget, here on El Reg, El Reg makes the rules and commentards can only help police them ...... which is a greater service than you are likely to be offered by many anywhere else in these days of hope and glory and real change ‽

Thanks for the assistance, John Smith 19. 'Tis much appreciated.

[anthonyhegedus]

There's no such thing as artificial intelligence. It's a simulation of a model of part of a model of how we think brains work, bolstered by lots and lots of statistics.

[Mage]

Re: model of how we think brains work

I don't think how we think real brains might work is anything to do with any computer system "Expert Systems", misnamed "Computer Neural Networks", misnamed "Machine Learning or misnamed "AI".

[holmegm]

I see

"Simply removing gender information was insufficient as the AI used other clues to prefer male applicants – because they were preferred in the data on which it was trained."

Does that roughly translate to "because they were more likely to have the relevant experience and qualifications"?

[Mage]

Re: more likely to have the relevant

No, because the young men got past HR. In the place I worked, women and older people only were permitted to be shortlisted for Engineering posts by HR if there was a shortage of applicants. Also they didn't get promoted if they did get in.

However there were less actual female applicants, because less females did engineering at college. Since then less are doing maths and computer science. So there are less female applicants.

It's not about their quality. Often above average because they needed to be to get that far.

[Alterhase]

Re: more likely to have the relevant

Mage wrote: Often above average because they needed to be to get that far.

When I think of the women in computer science, I think of Ada Lovelace and all her successors, including Grace Hopper, plus the "computers" who did so much during World War II and the space programs.

One of the most intelligent managers that I had the opportunity to work for was a woman.....

-- And why is there only one woman among the 30 icons offered to commenters, and why is that one woman Paris Hilton?

[holmegm]

Re: more likely to have the relevant

"One of the most intelligent managers that I had the opportunity to work for was a woman....."

I have had (and currently have) wonderful women managers. I have had few woman programmer colleagues.

[Adelio]

AI (Not inteligent)

All this talk about AI seems to be misleading. What wer are currently talking about at best is pattern matching. there is no inteigences with the current AI.

I think to say that the current AI are as smart as an earth worm is to denigrate earth worms.

We should STOP using the term AI for something as simplistic as what we can currently do.

ML or PM (Pattern matching) maybe but NOT AI.