anddt

Invisible Cities by Italo Calvino.

I guess you scrape odds from bookmakers' websites (I expecxt a fair bit of headless Selenium). Is there any legal implication of scraping data this way and reselling re-packaged as an API?

This, given the current state of things I belive that AI can be used, at most, as a more engaging/interactive piece of material to be used with the supervision of an actual teacher. The conversation that a student might have with GPT3 model will be coherent most of the time (and probably factually correct as well) but supervision from a teacher will always be needed to catch those cases where incorrect info is returned by the model.

I still believe that a bunch of AI-enabled edtech tools will bloom out in the foreseeable future, just nowhere near anything usable by the student withouth supervision.

Yes, both machines are not too recent (by now I'd expect to find them for cheap on ebay) but they support my workloads more than well. I work in ML and they're fine for general Python programming and exploratory data analysis. I spin a VM for anything that requires more memory/threads.

I'd would probably buy a beefier laptop from Tuxedo Computers or Slimbook if I needed more performances hoping to win the Clevo/Tongfang QC lottery (my T480s was fairly expensive in Italy 3/4 years ago).

So I guess that it ultimately boils down to the type of workload that OP deals with.

edit: poor grammar

Currently using a T490s (win-based on WSL, work machine) and a T480s (Fedora, upgraded to 24gb ram) for personal projects and they're both fine machines. The T480s runs a bit cooler after a repaste than the T490s but I'd say they can handle fine whatever I throw at them.

I can definitely relate to the way you described yourself and have been in a very similar position in the past. What helped is to chunk whatever I am doing in smaller chunks, for example getting to know the very basics when learning a new technology before diving into the more complex stuff and build upon that. This had the nice side-effect of multiplying the number of gratification moments.

Also, I've noticed I learn best when I actively try to limit the number of times I switch context within a day, this might require you to plan whatever it is you want to deep-dive beforehand, but your mileage might vary.

I worked as a contractor for a large media company in Europe (fully remote, from Italy). We had company issued laptops prepackaged with corporate tooling (VPNs, accounts, etc.) and that came with a fair bit of corporate-spyware included from _at least_ couple different vendors.

At one point, I was writing a small demo in golang for one of our projects and I've been contacted by a security engineer telling me that I've been hitting C:\Users\<yada>\AppData\Local\Temp\go-build2923888066\b001\exe\main.exe too frequently and that called a `cryptsp.dll` that according to him was highly correlated with ransomware attacks. I was adviced to stop working on that until my manager confirmed this was legitimate activity. I must admit, I've been quite freaked by the fact that they were listening for the single executables launched on my machine.

Needless to say, this dragged on for a week due to complex internal politics. I thoroughly enjoyed a week of paid time off.

> [...] Government statisticians tend away from black box AI and towards more traditional methods.

I was indeed wondering how constraints on model interpretation forces governments to stay away from NN and more complex approaches.

I have worked for a fraud-prevention company for a few years, and the financial nature of the decisions involved led to reward model interpretation (sometime at accuracy's expense) in order not to create friction with clients. Definitely not the best not being able to explain why your model performs poorly or what actions are needed to solve a problem.

I wonder what are the implications and liabilities to leave such a task to a private company. Let's say Sadiot overlooks some model behavior that leads to racial discrimination, who'd be liable? The government's data science team who set up the model or Saidot?

> no decision that has legal ramifications can be made without manual intervention

I think there's no other way around this: ultimately, everything that has legal implication will be potentially subject to dispute. We're not yet in the dystopian environment where disputes can be handled by a model :)

Sorry if it didn't come across in the post, it'd be no doubt a fair step towards more ethically sound governments.

I tend to believe this is a valid approach for ML application in public institutions. Being ML a black box or a potential source of discrimination, setting up systems to ensure credibility within the public should be addressed early on in the design process (perhaps while designing the model itself).

I wouldn't be surprised to see a future where all ML applications affecting citizens and the general public interest (excluding defense, military, etc.) will be obliged to be open-source and subject to public scrutiny.