A.I. Imitators and A.I. Problem Solvers – TMO Daily Observations 2023-01-11
On a day that Apple trumpets the many successes of its Services segment, ripoff ChatGPT apps are infesting Apple’s App Store. TMO Managing Editor Jeff Butts joins Ken to discuss the issue. Plus – generative A.I. isn’t just for drawing pictures and cheating on term papers anymore. Jeff looks at ways A.I. is building the cures of tomorrow.
Sponsors
Get rid of useless subscriptions with Rocket Money now. Go to rocketmoney.com/macobserver.
Ken and Jeff:
Enjoyed your discussion about AI and protein constructs.
Apparently, my comments were unpersuasive in allaying Jeff’s concerns. This is too complex a topic to sort a in podcast comment, so here goes.
Just because the shape and other properties of a molecule (eg a protein) indicate that it should behave in a certain way, and even if, using simple systems like an in vitro model (think Petri dish) or worse, and AI algorithm (this is what the computer said it would do), it does the thing we predicted it would, this DOES NOT mean that, once introduced into a complex system (think any mammal) it will behave as we expected.
It does not matter who is working on the molecule, whether in academia or private industry (let’s leave military weaponry aside for a second), typically a combination of both, experimentation with said molecule must pass through institutional review (especially in academia), and if it is to be used in humans, a well-described sequence of animal and human studies, the latter being publicly registered, and reviewed not only by an independent data safety monitoring board (DSMB) but by, certainly in the USA, the FDA before it can be approved for human use.
Case in point, Moderna’s most recent bivalent COVID-19 booster vaccine, despite having passed a tried and true method known as a ‘bridging study’ that should predict how effective a vaccine should be in creating immunity, under-performed in real humans https://www.cnn.com/2023/01/11/health/moderna-bivalent-transparency – a fact that Moderna was less than candid in sharing with regulators. This is a great way, BTW, of getting oneself into deep poop with regulators and getting a product or a licence pulled from the market. In this case, the ‘molecules’ (antibodies) simply failed to meet expectations against the two common Omicron variants against which they were targeted, but did not cause any harm. Still, this system prevents not only reckless experimentation, but product release onto populations, without extensive testing under safety protocols, and realtime proactive review by multiple independent bodies – at least in free societies (okay, I’ve probably just caused Jeff another sleepless night). Obviously, there is much more than can, and should be, said on this topic for fair and adequate treatment, but that’s the stick figure sketch.
As for weaponisation, the major powers are signatories to conventions against the use of bio weapons. That’s a topic for another time.
Cheers.