This was exactly my experience. We used DDS for our non-ROS parts but ROS2 was experimental at the time so we ended up needing to build bridges for everything that was ROS. Brining in even a single ROS package to not have to reinvent things was a huge pain point and ROS had issues where it couldn't be used across everything.
Location: NY, USA
Remote: Yes
Willing to relocate: Generally no but maybe
Technologies: Rust, Python, C/C++, PyTorch, NumPy, Docker, Bash, Linux, ROS[1]
Résumé/CV: https://resume.strom.ai/derek_goddeau_cv.pdf
Email: see résumé
Software engineer with strong experience in embedded systems, autonomous systems, distributed systems in very unreliable environments, and computer vision.
May also be interested in opportunities outside of those areas.
These things still feel a bit like e.g. Google/GCP services to me: Super appealing at first glance, quite close to what you want, but somehow never quite there. Maybe they'll asymptotically get there, eventually? Perhaps that statistical model can't really make it to the level we want it to?
I’ve found that replacing the bad parts with new ones, like Dalle Outpainting, can remove the worst parts of the image, like the hands here… doesn’t make it perfect, but certainly removes the worst offenders that instantly bring attention to themselves.
It may be that it's the deep learning tech which will never quite get there. GPT-3 has similar shortcomings in its mimicry. We're 95% there, I guess, but may never quite reach 100%.
Nah, the current issues are just because we're trying to do everything in one step. Because we've built tools that have so much of a stimulus-response approach, few efforts have been made toward interfaces that ask for clarification ('when you say X, do you mean XYZ or XXX?').
Image-to-image and tuning already addresses many of these issues; just as inpainting works really well, it won't be long before we have select-and-repair, where you add an additional prompt like 'improve this part - the ice cream is fine, just work on the dog's muzzle.'
The mistakes the AI makes are too numerous and hard-to-define for this to work I think. They could perhaps be addressed by having two different models trained differently, each fixing the errors of the other. When humans draw a realistic artwork, it's not 'single-pass'; they have to iterate on the details to get it right.
I get the same feeling as well. This approach may well be eternal demo-ware, and you'll actually need AGI (or manual direction by a real human) to get to 100%.
The hands throw me off. The same with the cat holding the remote... never thought that hands on animals would be able to trigger my uncanny valley response, but here we are
if people weren’t so repressed, this could also be used to severely reduce exploitation in the porn industry. what’s the point in making and selling exploitative porn when it can be auto-generated at will?
GitHub should not be sued for training on the data, but anyone using it should be liable for any copyright infringements it generates. That would effectively make it useless for business use cases, but it should be until the models understand copyright and plagiarism, which they do not yet.
If Microsoft asserts and represents that their tool doesn’t generate copyright-infringing code, then surely Microsoft is the party which should be liable, rather than the poor unlucky programmer who was lied to by the billion-dollar corporation’s marketing agency?
> surely Microsoft is the party which should be liable
unless microsoft is doing work-for-hire for you via copilot, i highly doubt they are liable.
You, as the person who is claiming to have produced the work (even though you were using a smart tool to help), must be the person who also is liable. Otherwise, could you not claim that the auto-correct on your word-processor is liable for copyright infringement?
Web development usually involves very little CS it is much closer to Software Engineering. Computer Science about solving problems with math, science, and computation theory and just happens to use computers as tools. Software Engineering is about building complete and useful programs.
I've encountered some "computer science" programs that barely taught people to program in Java, included classes on Microsoft Excel and Access and didn't include any advanced math. Data structures and algorithms might get a single class to satisfy interview questions about linked lists vs arrays and what Big O notation means. It's kind of a polluted term I think, in part due to that kind of usage by community colleges and the like.
Definitely switch to zsh for interactive use, the differences are minimal enough it will not take any effort to switch. Even forgetting everything else just the superior tab complete is enough to make it worthwhile. Still use bash for scripts they will be more portable.
