Artificial Intelligence #169

Hey, in this issue: ChatGPT outperforms crowd-workers for text-annotation tasks; the stickle-brick approach to big AI; a tutorial on finetuning LLMs on a single GPU using gradient accumulation; ChatGPT is polluting the web; and more.

This issue is sponsored by?Verta?and?Pachyderm.

• [IEEE]?The stickle-brick approach to big AI
• [Tutorial]?Finetuning LLMs on a single GPU?using gradient accumulation
• [ArXiv]?ChatGPT outperforms crowd-workers?for text-annotation tasks
• [Tutorial]?Logging for ML model deployments
• [Sponsored] Verta |?Try Verta Model Catalog free?to manage all your ML models

Join this 30-minute event?to see how easy it is to manage your ML models on the industry’s leading platform.?Verta Model Catalog centrally manages all model assets?with integrated deployment, versioning & monitoring to establish governance and enable collaboration and visibility in model performance.

• [Microsoft] SynapseML:?an open-source machine learning framework for distributed ML on Apache Spark
• [Data] A collection of?open-source datasets to train instruction-following LLMs
• [Sponsored]?Completing machine learning loop
• ‘As an AI language model’:?the phrase that shows how AI is polluting the web
• Stability AI launches the first of its?StableLM suite of language models

More than 650,000 subscribers are reading this newsletter. If you are building an AI or a data product or service, you can become a sponsor of one of the future newsletter issues and get your business featured in the newsletter. Feel free to contact?hello@truepositive.ca?for more details on sponsorships.

Enjoy the newsletter? Help us make it bigger and better by sharing it with colleagues and friends.