The definitive weekly newsletter on A.I. and Deep Learning, published by Waikit Lau and Arthur Chan. Our background spans MIT, CMU, Bessemer Venture Partners, Nuance, BBN, etc. Every week, we curate and analyze the most relevant and impactful developments in A.I.
We also run Facebook’s most active A.I. group with 191,000+ members and host a weekly “office hour” on YouTube.
Editorial
Thoughts From Your Humble Curators
We cover Paige.ai and what’s inside Google Dataset Search this issue.
As always, if you find this newsletter useful, feel free to share it with your friends/colleagues.
This newsletter is published by Waikit Lau and Arthur Chan. We also run Facebook’s most active A.I. group with 175,000+ members and host an occasional “office hour” on YouTube. To help defray our publishing costs, you may donate via link. Or you can donate by sending Eth to this address: 0xEB44F762c58Da2200957b5cc2C04473F609eAA65.
Join our community for real-time discussions here – Expertify
News
The Story of Paige.ai
We have a two-week break since last issue. So this piece about Paige.ai is around 10 days ago. But the story of Paige.ai is eerily similar to the DeepMind-Royal Free event which we have covered extensively on Issue 20 and 42. So we decide to a closer look.
Someone in Sloan Kettering’s Cancer Center decide to monetize the historical data and notes amassed over 60 years. More importantly, the company has exclusive use of the center’s 25 million tissue archive slides. That data corpus is VERY valuable.
The problem is none of these was disclosed to the staff pathologists beforehand. And the center’s executives own stakes of the new startup, Paige.ai. There are also potential privacy-related questions, even if the data have been anonymized (how anonymized?)? When all this came to light, it caused an uproar at the center. Since Sloan Kettering is a non-profit and Paige.ai is for-profit, the optics do not look good.
The NYT story gives a nice account of what happened. We recommend you to check it out.
We’ll leave you with the words of warning issued by the independent commission office (ICO) which investigated the DeepMind-Royal Free affair: Just that you can, it doesn’t mean you should.
Blog Posts
Inside Google Dataset Search
Google Dataset Search (GDS) has become our go-to website to search for database since its inception. When you think about it, building GDS, has to be a challenge in and of itself. How do you gather unstructured data from the web? How do you deal with replicated data? How do you gather this information and how do you structure them in a data structure.
All these questions are answered by this interesting post from the Google AI Blog. No surprise, none of these questions are trivial. For example, on the issue of representing the data, the researchers try to reconcile with the underlying Google Knowledge Graph (GKG) structures.
The Best Machine Learning Books
Normally we are not into sharing lists of resources. Well, it’s just cliche.
But when Uri Eliabayev, an AI consultant, as well as the the administrator of Machine and Deep Learning Israel (MDLI) showed us his list, we paid attention. Uri’s list are gathered from experienced members of MDLI and all of them are gems. We recognize several gems: PRML? Duda and Hart? And Goodfellow’s Deep Learning. These are books which help us in the past as well, and we found them not only useful references, but must-read textbook to improve you ML knowledge.
RTX 2080 Ti Benchmark
Lambda Lab did an interesting benchmark on how RTX 2080 Ti match up with its predecessor. TL:DR, the gain is truly impressive. Check out their link if you are thinking of upgrading your GPU card.
Video
Human-centric SDC
MIT Researcher Lex Friedman shared this interesting link of his latest work on human-centric SDC. We found it quite impressive because unlike most current-day SDCs which focus on autonomous driving, Friedman’s put more emphasis on human’s emotional state and how that affects driving. This thinking of human in the loop of AI is not new, but Friedman perfectly showcase in his video. So check it out!
Member’s Question
Ask AIDL: Why People Choose to do AI
As answered by AIDL members. A little light-hearted, yet genuine.
About Us
Join our community for real-time discussions here: Expertify