Veteran journalist and distinguished blogger, Robert X Cringely, is back from a multi-year hiatus. His first couple of posts were that he is back and where he had been. The latter leads into his current post on the AI industry's betting on failed models.
Because that is what the essay became, whatever Amodei intended. It gave every other person writing nine- and ten-figure checks a reason not to worry about the one thing that should worry them most. The hallucination problem is the difference between a clever toy and a system a hospital or a bank or a court can actually rely on. It is the whole ballgame for enterprise AI. And the prevailing wisdom, blessed from the top, is that you needn't address it directly. Scale will provide.
Look at where the money is going and you can see the permission slip being cashed. Stargate, half a trillion dollars. The hyperscalers, tens of billions each per year. The Anthropic–Akamai arrangement, nearly two billion more. The collective bet of the wealthiest companies in the world is that you fix intelligence — including its honesty — by buying more of it. The data center operators are happy. The chip vendors are ecstatic. The labs raising money at valuations with too many zeros are happy. Everyone in that chain has the same incentive, which is to believe that the answer is more.
The customers who will eventually pay for all of it are the ones who should be asking whether any of this is true.
In his previous post, he covered his first hand experience with LLMs not being fit-for-purpose in their very design.
Previously:
(2020) Bob Cringely is Still Alive and Kicking
« Researchers Develop A New Process To Get Lithium Out Of Rocks | ‘AI Washing’: Firms Are Scrambling to Rebrand Themselves as Tech-Focused »
Related Stories
Bob Cringely is running a new prediction at https://www.cringely.com/2020/03/25/2020-brings-the-death-of-it/
2020 Brings the Death of IT
IT — Information Technology — grew out of something we called MIS — Management Information Systems — but both meant a kid in a white shirt who brought you a new keyboard when yours broke. Well, the kid is now gone, sent home with everyone else, and that kid isn't coming back... ever. IT is near death, fading by the day. But don't blame COVID-19 because the death of IT was inevitable. This novel coronavirus just made it happen a little quicker.
I mentioned the switch from MIS to IT because that name change presaged the events I am describing here. Management Information Systems was an artifact of big business, where corporate life was managed rather than lived. Information Technology happened when MIS escaped into the wild. MIS meant office buildings and Local Area Networks while IT includes home workers in their pajamas which, frankly, describes me at this precise moment.
To quote the immortal Al Mandel (why am I the only one who ever quotes the immortal Al?) "the step after ubiquity is invisibility." IT was the last visible vestige of MIS and now it, too, is gone.
Another recent post has an interesting view of venture capital: https://www.cringely.com/2020/03/25/prediction-covid-19-will-kill-a-ton-of-startups-or-so-it-will-seem-as-vcs-pull-back/
(Score: 5, Insightful) by Anonymous Coward on Saturday May 30, @10:43PM (2 children)
It's statistical inference. The results that you're getting are more strongly statistically associated with what you put in than the results that you would like.
What's the fix? More strongly associating your desired output with your given input, or breaking the statistical correlation between the incorrect output and the given input.
Damn, that's a bitch of a problem. Good luck with that. It's almost like you'd need.. a system that can think about things, and give a considered, reasoned conclusion. You know - rather than a statistical correlation. (Which is to say: this is an intractable problem until true Artificial Intelligence comes online. Until then, you're trying to break every non-truth that the internet has seen, or create more truth, with every possible association, than there has ever been untruth on the internet.)
(Score: 1, Informative) by Anonymous Coward on Sunday May 31, @03:42AM
I've enjoyed reading Cringely since InfoWorld days, good entertainment, often a good laugh. But I don't put much stock in his comments, and less in his attempts at products or other commercial ventures.
Also, wasn't this pen name used by two different people?
(Score: -1, Troll) by Anonymous Coward on Sunday May 31, @07:21PM
There's a phenomenon in the world where, although bad actors exist, they are outnumbered in most parts of society. So, if you run a statistical sampling of actors (particularly a statistic like median), you are highly likely to find good actors in your output.
On the internet in general, it's true that there are areas where more bad / faulty information is "out there" than good information. Simple statistical inferrence doesn't cut it that way. However, if you break things down into their constituent components and those components are simple / straightforward / repeatable enough that they can be reliably statistically inferred, validated from various perspecticves, build a database of "core truth" and start evaluating higher level sources on the topics based on how self-consistent they are, how well they agree with "ground truth," then you can start to more accurately infer "truth" for more complex questions.
If a claimed cure for a disease is presented nakedly with only scant studies from sources with conflicts of interest to back up its efficacy, that's a weak claim. If that same claim demonstrates the cure utilizes biochemical mechanisms which are widely known to be true, and reasonably linked to the claimed result, that strenthens the claim.