This seems really cool for a company that has enough surface area of software where chances are somebody has already built the thing you need and instead of building a copy yourself, you can integrate with the existing system
Launch HN: CodeComplete (YC W23) – Copilot for Enterprise
51–60 of 69 posts
Re: Launch HN: CodeComplete (YC W23) – Copilot for Enterprise
#52> Permissively Licensed: Trained only on permissively-licensed repos to avoid legal risks You’re fine-tuning the model. What model are you fine-tuning? I can’t imagine you trained your own LLM from scratch, so how can you possibly guarantee the core model wasn’t trained on non-permissively licensed code?
We're starting off with an open source base model that was trained on The Stack, a dataset containing only permissively-licensed code, and we're further augmenting it with additional repositories with MIT, BSD, or Apache Licenses.
How are you complying with the attribution requirement for MIT and Apache?
From the MIT License[0]:
>The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
From the Apache License[1]:
>You must give any other recipients of the Work or Derivative Works a copy of this License; and You must cause any modified files to carry prominent notices stating that You changed the files; and You must retain, in the Source form of any Derivative Works that You distribute, all copyright, patent, trademark, and attribution notices from the Source form of the Work, excluding those notices that do not pertain to any part of the Derivative Works; and
Re: Launch HN: CodeComplete (YC W23) – Copilot for Enterprise
#53This seems really cool for a company that has enough surface area of software where chances are somebody has already built the thing you need and instead of building a copy yourself, you can integrate with the existing system
Code discovery and search have been great applications of Copilot and other AI dev tools!
Re: Launch HN: CodeComplete (YC W23) – Copilot for Enterprise
#54Excited to try this looks really good! Been looking for something like this for our Pynecone code base!
Let's talk :). Congrats on the Pynecone launch!
Re: Launch HN: CodeComplete (YC W23) – Copilot for Enterprise
#55Re: Launch HN: CodeComplete (YC W23) – Copilot for Enterprise
#56Earlier quoted context omitted.
Wasn't copilot only trained on public GitHub repos? Presumably there's nothing stopping anyone from training on them.
Yea, there are tons of public data to train on. Copilot's under fire right now: https://githubcopilotlitigation.com/ , but we make sure we only train on permissively licenses because some companies are sensitive to the IP issues here.
Can you give a list of the permissive licenses you train on? MIT for example requires attribution. In fact, most permissive licenses have similar requirements.
Re: Launch HN: CodeComplete (YC W23) – Copilot for Enterprise
#57Earlier quoted context omitted.
Yea, there are tons of public data to train on. Copilot's under fire right now: https://githubcopilotlitigation.com/ , but we make sure we only train on permissively licenses because some companies are sensitive to the IP issues here.
Can you give a list of the permissive licenses you train on? MIT for example requires attribution. In fact, most permissive licenses have similar requirements.
MIT only needs attribution for redistributing "all copies or substantial portions of the Software".
Juat reading the code shouldn't be illegal, and producing a short snippet based on the code but not identical shouldn't count as a "substantial portion", but IANAL...
Re: Launch HN: CodeComplete (YC W23) – Copilot for Enterprise
#58I think this a problem that is worth solving as not only the biggest companies are concerned with security, many SaaS and tech companies are concerned about source code of product become compromised.
Re: Launch HN: CodeComplete (YC W23) – Copilot for Enterprise
#59GitHub and other companies like Amazon have the advantage of the scale in terms of dataset. What’s the guarantee that the pre trained model you have that you’ll fine tune on a company’s code base is as good as say Copilot? It makes it even hard to evaluate when you don’t have a demo to try - it’s not that hard to setup a pipeline to run your model in cloud and send invites to potential customers if you want to.
It doesn't make sense to treat Copilot as a competitive benchmark where Copilot isn't even an option for policy reasons, which is what this product is targeting. Does it provide net positive support to developers? Is that support worth the licensing and maintenance costs? Those would indeed benefit from a demo, but it's also hard to demo something whose value hinges on fine-tuning. More likely, they'll need to sell l…
What compliance issues are there? Copilot is already being adopted at the enterprise level. Microsoft is not going to miss out on enterprise by trying to ship data out.
Re: Launch HN: CodeComplete (YC W23) – Copilot for Enterprise
#60Always astonishes me how negative hacker news can be whenever people try to launch stuff. This is cool!
Because since there isn't any public demo available, you can't really say more than "Yay, this seems interesting."
So it's easier to point out obvious potential flaws.