Decoding the Genome: Unraveling the Complexities with AI and Creativity
ฝัง
- เผยแพร่เมื่อ 5 มิ.ย. 2024
- Support us! / mlst
MLST Discord: / discord
Twitter: / mlstreettalk
In this eye-opening discussion between Tim Scarfe and Prof. Jim Hughes, a professor of gene regulation at Oxford University, they explore the intersection of creativity, genomics, and artificial intelligence. Prof. Hughes brings his expertise in genomics and insights from his interdisciplinary research group, which includes machine learning experts, mathematicians, and molecular biologists.
The conversation begins with an overview of Prof. Hughes' background and the importance of creativity in scientific research. They delve into the challenges of unlocking the secrets of the human genome and how machine learning, specifically convolutional neural networks, can assist in decoding genome function.
As they discuss validation and interpretability concerns in machine learning, they acknowledge the need for experimental tests and ponder the complex nature of understanding the basic code of life. They touch upon the fascinating world of morphogenesis and emergence, considering the potential crossovers into AI and their implications for self-repairing systems in medicine.
Examining the ethical and regulatory aspects of genomics and AI, the duo explores the implications of having access to someone's genome, the potential to predict traits or diseases, and the role of AI in understanding complex genetic signals. They also consider the challenges of keeping up with the rapidly expanding body of scientific research and the pressures faced by researchers in academia.
To wrap up the discussion, Tim and Prof. Hughes shed light on the significance of creativity and diversity in scientific research, emphasizing the need for divergent processes and diverse perspectives to foster innovation and avoid consensus-driven convergence.
Filmed at www.creativemachine.io/
Prof. Jim Hughes: www.rdm.ox.ac.uk/people/jim-h...
Dr. Tim Scarfe: xrai.glass/
Pod: podcasters.spotify.com/pod/sh...
Table of Contents:
1. [0:00:00] Introduction and Prof. Jim Hughes' background
2. [0:02:48] Creativity and its role in science
3. [0:07:13] Challenges in understanding the human genome
4. [0:13:20] Using convolutional neural networks to decode genome function
5. [0:15:32] Validation and interpretability concerns in machine learning
6. [0:17:56] Challenges in understanding the basic code of life
7. [0:19:36] Morphogenesis, emergence, and potential crossovers into AI
8. [0:21:38] Ethics and regulation in genomics and AI
9. [0:23:30] The role of AI in understanding and managing genetic risks
10. [0:32:37] Creativity and diversity in scientific research - วิทยาศาสตร์และเทคโนโลยี
As an ex molecular biology/bioinformatics grad who is currently work with ML, this tape is a delight. I really love that you recently started explore other fields and their relationship with ML.
Ey up you're a busy bee today! Loved the debate with Connor and looking forward to watching this one. I know it's cliche but cheers for the graft. There are people who bloody appreciate the effort that you and others like you are putting in. I'm stuck working a 9-5 but I fully support anyone who is flying the flag of all this cool ai stuff!
Seconded!
Notice how the real conversation starts when you think it’s all over …he got real and less robotic and expressed the meat of the podcast ….this is way long form and relaxed the magic comes out
The interviewer knows how to drive a very interesting conversation
Haha, pretty cool. Enjoyed this conversation. Tim, carry on with letting your curiosity lead you down these seemingly divergent paths. It is greater creativity we need at the moment in the field of Deep Learning and I feel that is where MLST's real contribution lies. In that spirit I would like to recommend you take a look at the work of Dr. Iain McGilchrist and Dr John Vervaeke both who I believe may have very interesting things to say both about the paths to GAI, also comment on the current developments with LLMs.
As someone pivoting back to academia after a decade of game design and circus arts, this was super inspiring to me. Does anyone have recommendations for how more 'creatives' can get involved in the field(s) of cog sci / AI / creativity research?
Personally I'd find it so fascinating to research how the concept of Play might reveal itself in AI; seems to me that play has a pretty core role in our cognitive development. Guess it might be time to head back to grad school o7
Keep em coming, im hooked!
I really enjoyed watching this one.
[Off topic, but...]
Hey, Tim !
Haven't you forgotten to release the full interview with Minqi Jiang ? He has a great combination of depth of knowledge and ability to explain, the full interview must be a gem! Don't hide it for yourself only :)
It went out on the podcast months ago podcasters.spotify.com/pod/show/machinelearningstreettalk/episodes/114---Secrets-of-Deep-Reinforcement-Learning-Minqi-Jiang-e22fkp7 -- if you are referring to releasing the full video here on TH-cam - I wasn't originally planning to as it's a huge 3 hour interview although I might release it in the near future if I get time to edit it
@@MachineLearningStreetTalk
A pity... I "live" exclusively on TH-cam.
I find that seeing the person who's talking helps me focus, thus I don't like "sound only" podcasts. And listening to a podcast while doing something else ? What kind of heresy is that ?! :)
I suppose your brand disallows releasing anything with minimal editing ? :)
Ok, good to know it's within reach.
Thank you for your response.
Keep up the good work ! :)
just like LLM trained on lots of text can produce consistent text, a large genome model trained on many genomes could produce entire genomes
💓
I never understand why people mix up diversity of thought with physical diversity (as in gender/race, obviously through is physical to some degree too). Diversity of ideas is so much better than lowering standards to meet physical diversity standards.
because diversity of cultures and experience usually equate to diversity of thought..... i don't think this is that complicated of an idea....
Truth is....most papers are useless.😂