[Red Badger]

[Red Badger]

Tech Ping!....................

[Joe Boucher]

Exactly.

[Red Badger]

[Jack Hammer]

Colossus: The Forbin Project

[M Kehoe]

Dave: Open the door HAL.

5.56mm

[OneWingedShark]

There's no way this could go wrong!
No way at all!

[WayneS]

Google’s New Computer With Human-Like Learning Abilities Will Program Itself

... to KILL you!

[null and void]

[Fred Hayek]

Skynet

[Steely Tom]

So why not just do it? Why talk about it first?

The Japanese “Fifth Generation” project back in the late 1980s made the same promise.

[Billthedrill]

The thing will never be able to call itself a programmer until:

1. It flies into a rage at "Bad Command Or Filename"

2. It chirps "I don't need to document, the code is commented."

3. It chirps "I don't need to comment the code, I know what I'm trying to do."

4. It growls "Change control is a complete waste of time."

5. It passes the Turing test with "Hello world"

6. It is capable of body odor.

7. It can type a three-line shell command with two fingers but can't type a shopping list without misspelling "mayonnaise"

8. It knows fewer that three other AIs that aren't programmers.

[adorno]

New Computer With Human-Like Learning Abilities

Yeah, "human like learning" of those with an IQ of 1.

[InterceptPoint]

Time for a somewhat serious post.

It was Kurtzweil and probably others who noted that technological progress is exponential not linear. This idea of Google’s seems like an excellent example of that proposition. The advantage that computers have over human beings is that they can easily transfer information between themselves with 100% accuracy. Another advantage they have is that they can work 24 hours a day 365 days a year. It seems like an obvious conclusion that these machines, as they learn to program themselves, will improve their capability over time in an exponential manner.

That may mean an acceleration of progress for the human race or our doom.

[backwoods-engineer]

This article is utter bullcrap and marketing. I worked for a company in 1988 that was using neural networks to learn how to process unknown data. At the time, it was under a FedGov contract, and they were trying to match certain communication signals.

Pairing neural networks and ordinary digital computers is NOT new.

[thorvaldr]

Sarah Connor: What did he just say?
Gas Station Attendant: He said there’s a storm coming in.
Sarah Connor: [sighs] I know.

[mikrofon]

They should call it- “Kaci” & make it a medical droid, since obviously you can acquire this knowledge extensively on your own……

[TBP]

I Think We're All Bozos on This bus.

[TBP]

I Think We're All Bozos on This Bus.

[Yollopoliuhqui]

AI - No history of psychological trauma, no subconscious mind causing Freudian slips, forgetfulness and bad decision making, no psychological attachments to objects or authority figures, no irrationality, no mood swings, no imaginary illnesses, no malingering, no labor union problems, no absenteeism...

What's not to like?

[rightwingcrazy]

I created self-modifying code using Forth, decades ago. I presume they’ve done something much cooler, being Google and all.