[FoRK] The AI breakthrough
meltsner at alum.mit.edu
Sun May 7 18:11:44 PDT 2017
The advantage of traditional AI -- rules, frames, etc. -- is that the
process to get to the results was comprehensible, and typically
defensible. We are now in a world where ML-based tools produce better
translations than dictionaries and rules, except when it goes
I agree that most of us had given up on true AI; some of us still had
hopes of understandable and explainable competence. That seems to be
at end -- the simpler cases have become part of the mainstream, the
harder ones have been tackled, usually successfully, by ML.
Would you want to send traffic over a novel bridge design produced by ML?
On Sun, May 7, 2017 at 6:02 PM, Lucas Gonze <lucas.gonze at gmail.com> wrote:
> Bill K is correct. This is just the post the mebot would generate.
> What the I-bot am thinking about is the Rubicon, as Aaron calls it. Though,
> it's not a specific application like speech recognition on the bot's mind,
> it's ML as a general tool.
> I wonder about comparable events in the history of software development.
> The birth of Unix? SQL? Transistors? Von Neuman?
> Dr Ernie, it's funny in a gruesome way to think of ML as invalidating all
> the decades of procedural AI research. My thought is that true AI has been
> a laughing stock all these years, with procedural methods sitting at the
> captain's table. Most of us - definitely including me - considered true AI
> a wild goose chase. I was aware of methods to approximate cognitive
> processes. I believed human intelligence could be rivaled by automated
> systems. But I didn't think it would matter in daily engineering except for
> edge cases.
> On Sun, May 7, 2017 at 5:15 PM, Bill Kearney <wkearney99 at hotmail.com> wrote:
>> Ok, so who here thinks/realizes this is more bot that Gonze?
>> AI, indeed...
>> -----Original Message-----
>> From: Lucas Gonze
>> Sent: Sunday, May 07, 2017 3:33 PM
>> To: Friends of Rohit Khare
>> Subject: [FoRK] The AI breakthrough
>> The mainstream-level breakthrough of AI / machine learning in the last 4-5
>> years is a watershed in computer science. It validates decades of research.
>> FoRK mailing list
> FoRK mailing list
After 30+ years of email, I have used up my supply of clever .sig material.
More information about the FoRK