Protein folding AI: "Will Change Everything"

Click For Summary

Discussion Overview

The discussion centers around the implications of Google's deep-learning program for predicting the 3D shapes of proteins, particularly its potential to transform the field of biology. Participants explore the nature of AI's role in biochemistry, contrasting it with human reasoning and the methodologies employed in protein folding predictions.

Discussion Character

  • Exploratory
  • Technical explanation
  • Debate/contested

Main Points Raised

  • Some participants suggest that AI could revolutionize biology by handling complexities beyond human cognitive capabilities.
  • Others argue that the AI's approach is more about inducing principles from a vast array of experimental results rather than deducing from fundamental principles.
  • A participant notes that while physical principles are utilized to some extent, the program primarily relies on eliminating physically impossible structures, indicating a hybrid approach of induction and physical calculations.
  • Links to external articles are provided to support various viewpoints on the achievements and implications of the AI program in protein folding.

Areas of Agreement / Disagreement

Participants express differing views on whether AI in this context is more about induction or deduction, and there is no consensus on the extent to which physical principles are integrated into the AI's methodology.

Contextual Notes

Some limitations include the unclear definitions of deduction and induction in this context, as well as the unresolved nature of how physical principles are applied within the AI program.

Biology news on Phys.org
Biology looks to be where AI can really shine, as it is simply too complex for human brains to grasp. Is the promise that AI can 'derive' biochemistry like a good human physicist can deduce a good hypothesis for physical phenomena from basic principles?
 
As far as I understand it is definitely not deducing from basic principles, it is more like deducing any principles from experimental results. A very large number of them, and more than unassisted human minds manage to see the underlying patterns in. Er, I think that's induction rather than deduction.

Physical principles are used, seemingly to a limited extent by the programme, according to this related article https://www.nature.com/articles/d41586-019-01357-6 - the programme runs through structures to eliminate physically impossible ones. I would guess the combination of approaches, using physical calculations to refine the inductively predicted structures will in future be very powerful.
 
  • Like
Likes   Reactions: BWV

Similar threads

  • · Replies 2 ·
Replies
2
Views
1K
Replies
3
Views
2K
  • · Replies 1 ·
Replies
1
Views
2K
  • · Replies 2 ·
Replies
2
Views
1K
Replies
10
Views
5K