- cross-posted to:
- [email protected]
- cross-posted to:
- [email protected]
Oil and gas companies need to integrate AI into their operations so that they can extract more fossil fuels—which can then be used to power more AI.
That is the pitch that Canada’s leading AI company, Cohere, is making to the oil and gas industry, even as its CEO Aidan Gomez argues publicly that it is important for the tech industry to be aligned with “environmental goals.”



Cohere is struggling to stay relevant within the LLM space. That’s why they’re making noise like this.
They used to have an interesting niche as “business RAG models” for mundane tasks (think mass document processing or retrieving stuff from documentation). The LLM tinkerer community liked that, as their models were open weights and also good for storytelling/dungeonmasters/roleplay or whatever (as they were tuned to use context instead of hallucinating chatbot slop).
Cohere stayed low key. They didn’t need to make noise.
But other open weights models have surpassed them in efficiency. I’m not sure why one would even “pick” Cohere anymore, unless being Canadian was important.
Being Canadian was important enough for the government to throw money their way:
Cohere doesn’t seem super ambitious, for an AI company:
The new data centre that’s supposed to be built in Nova Scotia for Anthropic is supposed to cost $8 billion.
If Cohere were smart, they wouldn’t use any of that infrastructure for training: just finetune other open weights models and optimize/specialize them for their clients.
And pool stuff with Mistral. Working with a French company would make a lot of sense, right?
…And this is what irks me. It looks like they’re taking the “Tech Bro” approach of overspending to ape OpenAI… rather than actually doing something practical.