
ANTHOLOGY — Open source AI (Interview)
Streams straight from the publisher. podnod never proxies or re-hosts episode audio.
This week on The Changelog we’re taking you to the hallway track of The Linux Foundation’s Open Source Summit North America 2023 in Vancouver, Canada. Today’s anthology episode features: Beyang Liu (Co-founder and CTO at Sourcegraph), Denny Lee (Developer Advocate at Databricks), and Stella Biderman (Executive Director and Head of Research at EleutherAI).
Special thanks to our friends at GitHub for sponsoring us to attend this conference as part of Maintainer Month.
Changelog++ members get a bonus 3 minutes at the end of this episode and zero ads. Join today!
Sponsors:
- DevCycle – Build better software with DevCycle. Feature flags, without the tech debt. DevCycle is a Feature Flag Management platform designed to help you build maintainable code at scale.
- Sentry – See the untested code causing errors - or whether it’s partially or fully covered - directly in your stack trace, so you can avoid similar errors from happening in the future. Use the code
CHANGELOGand get the team plan free for three months. - Rocky Linux – Enterprise Linux, the open source community way.
Featuring:
- Beyang Liu – GitHub, X
- Denny Lee – GitHub, LinkedIn, Mastodon, X
- Stella Biderman – Website, GitHub, LinkedIn, X
- Adam Stacoviak – Website, GitHub, LinkedIn, Mastodon, X
- Jerod Santo – Website, GitHub, LinkedIn, Mastodon, X
Show Notes:
The common denominator for these conversations is open source AI.
Beyang Liu and his team at Sourcegraph are focused on enabling more developers to understand code and their approach to a completely open source, model agnostic, coding assistant called Cody has significant interest from us.
Denny Lee and the team at Databricks recently released Dolly 2.0, the first open source, instruction-following LLM, that has been fine-tuned on a human-generated instruction dataset and is licensed for research and commercial use. They want to be the platform of choice the future of AI development.
Stella Biderman gave the keynote address on generative AI at the conference and works at the base layer doing open source research, model training, and AI ethics. Stella trained the EleutherAI pythia model family that Databricks’ used to create Dolly - 2.0.
- Cody from Sourcegraph - Read, write, and understand code 10x faster with AI. Cody answers code questions and writes code for you by reading your entire codebase and the code graph.
- Free Dolly: Introducing the World’s First Truly Open Instruction-Tuned LLM
- EleutherAI - Empowering Open Source Artificial Intelligence Research
Something missing or broken? PRs welcome!
Maintainer Month
maintainermonth.github.comJoin the discussion
changelog.zulipchat.comChangelog++
changelog.comDevCycle
devcycle.comSentry
sentry.ioRocky Linux
rockylinux.orgGitHub
github.comX
x.comGitHub
github.comLinkedIn
linkedin.comMastodon
hacky.townX
x.comWebsite
stellabiderman.comGitHub
github.comLinkedIn
linkedin.comX
x.comWebsite
adamstacoviak.comGitHub
github.comLinkedIn
linkedin.comMastodon
changelog.socialX
x.comWebsite
jerodsanto.netGitHub
github.comLinkedIn
linkedin.comMastodon
changelog.socialX
x.comCody from Sourcegraph
about.sourcegraph.comEleutherAI
eleuther.aiPRs welcome!
github.com
- 0:00This week on The Changelog
- 1:44Sponsor: DevCycle
- 4:38Start the show!
- 5:46We met Beyang 10 years ago!
- 6:08The mission of Sourcegraph
- 7:22Adam still Googles, just less
- 8:30Plugins make models interesting
- 9:35When did you start thinking about this?
- 12:16This is a "Eureka!" momement in time
- 13:11The gospel of text based input
- 15:44Is this the future interface of Sourcegraph?
- 17:21Iterating the interface
- 17:59How can you access Cody?
- 18:27Cody is open source
- 20:13How does it get code intelligence?
- 21:58What about privacy?
- 26:11GPT for X
- 26:53Cody vs Copilot
- 29:25Open source + model agnostic
- 31:22What's next?
- 33:19How high up the stack can AI tooling go?
- 36:07Is this a step change to plateau?
- 38:21The ultimate flattener
- 42:56Will AI awallow all of programing?
- 45:52Sponsor: Sentry
- 50:08We're fine-tuned
- 50:51JIT conference presenter
- 52:32This time 4 weeks ago
- 53:54Let's generate our own data
- 55:05All 15,000 Q&A data is open
- 56:12Verbose is not always desirable
- 56:42I want my own Dolly 2.0
- 58:14How did you collect the Q&A data?
- 1:00:39We thought we'd need more data
- 1:01:40Dolly proved it could be done
- 1:03:24Google's leaked memo
- 1:06:06Databricks' play in this chess game
- 1:08:45Turning AI on our transcripts
- 1:11:03Chain or foundational model?
- 1:12:42Sponsor: Rocky Linux
- 1:15:19The base layer
- 1:16:27What should the world know?
- 1:17:40Where does the money come from?
- 1:18:13Training LLMs is NOT that expensive
- 1:22:07Focused on open source AI research
- 1:25:49Interpreting LLMs
- 1:28:30Influencing the properties of the model
- 1:31:40Do you have fear of where this is going?
- 1:32:58Connecting with Stella and team
- 1:34:07Stella's news source is their Discord server
- 1:36:22Outro