HN user

BayezLyfe

24 karma
Posts11
Comments23
View on HN

Causalis | Israel, EU, US, Remote | Full-Time | https://causalis.ai

Causalis is a VC-backed startup partnering with the world’s leading pharma companies, building machine learning solutions to bring causal intelligence to healthcare. We are developing a platform of solutions based on novel and proprietary causal AI techniques, to help doctors, patients, and researchers understand and act on the cause-effect relationships in medical data. We are a multi-disciplinary, international team of experts in AI and ML, causal inference and data science, healthcare and medicine. We are looking for smart, driven and overall extraordinary individuals who are passionate about changing the world of personalized medicine, and eager to shape the future of Causalis -- https://causalis.ai/careers.

With new funding and pharma contracts, we're HIRING ENGINEERS #1 AND #2 in the following areas:

-> Infrastructure (maybe VP of Eng): https://causalis.ai/_careers-infra

-> Data Science + Data Engineering: https://causalis.ai/_careers-ds

Please email us your resume/CV (and reference this post): hello@causalis.ai

As a legitimate AI research engineer, I 100% agree with the take of the OP (particularly with the example self-proclaimed "Forbes AI Innovator of the Year"): I'm not salty because I'm jealous... I want more people to learn ML and Data Science, I just don't want them to learn snake oil selling. I'm particularly salty because being a snake oil salesman and a shameless self-promoter seems to be a legitimate path to success. As an academic and a scientist, it bothers me that people listen to advice from such snake oil salesmen.

Good points, but a back-of-envelope calculation looks like:

P(failure) = P(infection) * P(critical) * (1-P(icu))

0.1 * 0.8 * 0.9 = 0.072

That infection probability is optimistic, considering what you mention above. That ICU availability probability is the situation we're facing in NYC.

Related: "In general a machine learning system is built and trained to optimize a specified target objective: classification accuracy in a spam filter or tumor diagnostic, efficiency in route planning or Amazon box packing. Unlike these precise performance metrics, the criteria of safety, trust, and nondiscrimination often cannot be completely quantified. How then can models be trained towards these auxiliary objectives?"

From "Interpreting AI Is More Than Black And White": https://www.forbes.com/sites/alexanderlavin/2019/06/17/beyon...

Thank you. Appreciate the advice and kind wishes.

For better or worse, I have been exploring acquihire for a few weeks already, with several interested parties. However those discussions have not included the current circumstance, where I personally would not be 100% involved post-acquisition. I imagine this could be a dealbreaker.

You need someone you really trust, who is 100% independent

The examples here are not 100% independent, given prior friendships or working relationships with the founders. Terms will stipulate that the assignment of an independent director be "mutually agreed upon by the other board members" (e.g. founder and investor).

Establishment alpha shares foobar seemingly to help startup founders, while also aiming to maximize the number of founders aligned w/ alpha's preferred process.

Here: alpha is Atlassian, foobar is this new M&A term sheet

Also: alpha is Y Combinator, foobar is the SAFE

Also: alpha is Y Combinator, foobar is startup school

...

"Underspecified and misaligned notions of interpretation impede progress towards the rigorous development of understandable, transparent, trusted AI systems."

Powerful thesis. Essentially these are the steps forward:

1. precise definitions of interpretation, yielding interp-metrics

2. develop and train models that optimize both prediction accuracy _and_ interp-metrics

3. voilà, aligned objectives yield trusted AI