AI 'is a year away from knowing more than all human experts', those startled experts predict
AI will be ready to score full marks on one of the world's most challenging knowledge tests branded Humanity's Last Exam (HLE) in a matter of months, developers claim.
HLE was set up by tech bosses to see just how intelligent their systems are and consists of 2,500 meticulously chosen questions, spanning around a hundred topics from rocket science and mythology to physiology.
Each one requires at least PhD levels of understanding and to achieve a score even close to 100 per cent would earn someone the title of a 'universal expert'.
 |
| Ed - a leading Net Zero expert |
8 comments:
The question is how to be rid of this clown and reverse course ... two different tasks.
AI doesn't do "I". It mainly does regurgitation and a bit of correlation. It does, however, undoubtedly do lies because, being unable to think, it will offer an answer based on its models of probability and present it as if it were factual.
As for "at least PhD levels of understanding", it is the sort of thing said by someone who has no idea of what a PhD involves.
It reminds me of some sorts of characteristically American twaddle such as "you have to have an IQ of 140 to be a genius". It makes me suspect that many Americans think "genius" means 'one of the cleverest boys in my High School class'.
Net Zero is too good for Ed. He is a net debit from the sum of human understanding
James - we seem to be stuck with him until the failure of Net Zero causes so much damage that even the BBC has to admit it.
dearieme - I'm comfortable with naming the "I" in AI as intelligence. It isn't human and it makes mistakes, but is already able to summarise many subjects and issues much more cogently than most humans. Trust may be the issue, we'll expect it to be infallible.
I have my doubts about the expertise of AI since it mainly processes data that is already known rather than organising investigations into securing new data.
However if 'Expert AI' tells the world (including Ed Miliband) that Net Zero is a pipedream and could not be achieved for 100 years (i.e. when fusion power is routinely available) would the politicians come to their senses?
{spoiler alert}
No.
But what is it summarising? The stuff it's been trained on. Who selected that and on what principles?
decnine - yes he is a hole in the fabric of human understanding.
dearieme - we could ask the same questions about human summaries. One thing humans seem to have that machines don't is the curiosity to examine the sources of summaries and look for better ones.
As we know, many people can be conventionally intelligent but remarkably incurious. Ed Miliband does not seem to have the curiosity to examine the weaknesses in Net Zero or his father's politics yet he is conventionally intelligent enough to be a senior Minister.
DJ - if expert AI does tell the world that Net Zero is a pipedream, then a major political problem could arise. We see hints of it already in the ease with which we can search for information about the defects of Net Zero. It isn't easy to see what democratic politics can do about that apart from what we already see - restrictions on free speech.
Post a Comment