-
You have previous on here for getting a hard-on for AI without actually understanding it. Don't be surprised when this irritates people who have some knowledge either of the computing field or the specific area where the AI/ML you're uncritically hyping is being applied. It's not because they're Luddites, it's because they care.
Have I? I don’t recall. It’s not my field of expertise, and given how far outside of most people’s fields of expertise it is I’d hope 1) media did a better job covering it critically, and 2) experts/critics did a better job of voicing their concerns. I uncritically repeated what multiple MSM said, you got me there, but to be irritated at me for being interested in learning more (hence posting it here, on a bicycle forum, for discussion) and not doing it well enough for you (I presume an expert), seems harsh.
I think we can all agree that I post here too much anyway. Shrug* I’ll tone it down.
Eseman-
For what its worth I'm working for a machine learning company helping them write a tech design for a new neural network product at the moment and I've been impressed at how patient they are with people who don't understand the tech and/or get roped in by the hype. I do suspect it's mostly bicycle forum machine learning experts who are getting grumpy about it.
-
I think we can all agree that I post here too much anyway.
Ech, don't listen to anybody who tells you that. Some of the worst gatekeepers on this site still write dozens of posts here ever day despite having moved to whole other continents.
Stonehedge
itsbruce
That article, although it's not the only one.
No, it bloody well hasn't solved that problem and no respectable science correspondent (sadly, most are not respectable) should take that claim at face value. Sadly, both the Times and the Guardian blindly accept DeepMind's shameless hype of (checks notes) their own commercial product.
AlphaFold is a significant advance, but it hasn't solved the problem. It isn't hard to find informed experts in the field who give a more measured assessment. Try this, for example. That one is particularly good because it not only salutes the scale of the advance but also goes into detail about why anybody should be suspicious of DeepMind's claims given their commercial interests, the obstacles to reproducing their methods and their actual behaviour (e.g. all the patents they've now applied for).
Sometimes bad reporting of science is entirely down to the naivety of poorly-educated reporters, sometimes self-interested distortion by the researchers (or their employers), sometimes both. In this case, it's both. Hassabis absolutely knows the definition of "decidable" in computing/logic, so he knows he's telling porkies when he says the problem is solved. It isn't even solved for most practical purposes (nowhere near), let alone definitively.
You have previous on here for getting a hard-on for AI without actually understanding it. Don't be surprised when this irritates people who have some knowledge either of the computing AI field or the specific area where the AI/ML you're uncritically hyping is being applied. It's not because they're Luddites, it's because they care.