From the Gen leads Marco and Alex: https://discourse.julialang.org/t/whats-the-difference-betwe...
710 karma · joined September 26, 2014
My path: undergrad and masters in MechE, learned a little machine learning with matlab in grad school, taught myself python and hacked at AI/ML side projects while working at a startup doing FEA for about a year, eventually got an internship with an ML startup... eight years later I'm Chief Scientist at a near-unicorn AI startup.
They'll retweet to get more follows.
They'll +1 and forget about this.
They'll brag about some lab-grown meat investment that is really 0.01% of their portfolio that's chock full of productivity apps and luxury home goods.
They'll broadcast "we invest in the boldest founders changing the world with science and tech innovation", then back yet another consumer goods startup that literally perpetuates the economic challenges this article suggests to fight.
2. Round-closing announcements trail the actual closing by several months. Not to mention the actual raising and diligence takes months (even more so for larger/later rounds). Thus the $ reported here does not factor in covid19 and the current economic crisis, as implied.
[1] https://www.ft.com/content/21b19010-3e9f-11e9-b896-fe36ec32a...
2. Don't speak just to speak. Empty comments are wasteful -- minimal value in the eyes of dimwits, and negative value to those that matter.
Point 2 is related to "Law 4: Always Say Less Than Necessary" of Greene's 48 Laws of Power [1]. It would be good to read the full book ;)
Perspective from the other side of the table: I can't speak for every company, but we've worked day and night for weeks to figure out a plan that puts employees first and gives us a fighting chance to emerge from this crisis. It may not feel like it, but damn we fight for every single person's job. Paycuts w/o layoffs is a win.
My thought experiments have been hand-wavy expected value calculations like this: P(failure) = P(infection) * P(critical) * (1-P(icu))
0.1 * 0.8 * 0.9 = 0.072
and cost of failure is, well, infinite.
That infection probability is optimistic, considering what you mention above. That ICU availability probability is the situation we're facing in NYC.
Oh I see you have a dot-ai domain. Very cool.
https://programmingisterrible.com/post/139222674273/write-co...
2. To some people levels matter for career growth and validation. Other people don't give two sh*ts. For all people looking to switch jobs (as we all inevitably do) the levels make a difference in recruiting, salary negotiations, etc.
https://abe-winter.github.io/plea's/help/2018/02/11/slack.ht...
There are exceptions, but this is largely true.
I would go a step further and suggest FAANG are easier environments for ML PhDs because they go into research groups like FAIR and Google Brain, where legitimate software engineering (and general professional practices) aren't required. In other companies -- i.e. ML startups -- those training wheels need to come off, quickly.
"No matter the definition, developing an AI system to be interpretable is typically challenging and ambiguous. It is often the case that a model or algorithm is too complex to understand or describe because its purpose is to model a complex hypothesis or navigate a high-dimensional space, a catch-22. Not to mention what is interpretable in one application may be useless in another."
"Underspecified and misaligned notions of interpretation impede progress towards the rigorous development of understandable, transparent, trusted AI systems."
[1] https://www.forbes.com/sites/alexanderlavin/2019/06/17/beyon...
How to learn ML? Do fastai + reading Daphne Koller's and Chris Bishop's books on PGMs + re-implementing a paper on Gaussian process classification + another paper on GNNs + ....
No. You will not get beyond copy-paste level without being comfortable with ML foundations. That doesn't mean you need to be able to prove variational inference bounds in your sleep, but you'll want to know why we need things like lower bounds for approximate inference.
2. Implement their methods from scratch (i.e. numpy not pytorch)
3. Experiment a bit, tweaking the models/algs to gain intuition
4. Repeat 1-3
No, you're generalizing the marketing department at IBM over a deeply passionate, hard-working, brilliant community of scientists and engineers.