Great question. The CPG data landscape is a fragmented mess. We're actually tackling this at arena (https://arena-ai.com) to build out AI foundation models for the industry. We have some tips and potentially tools that might be useful to you. Would be great to get your input / hear your pain points if you're up for it
The subluminal application is particularly interesting, because it seems almost feasible. Is there a practical path to meeting these mass/energy requirements though?
Arena | Lead Machine Learning Scientist + Deep Learning Scientist | Full-time | Remote / NYC
—
We founded Arena a little over a year ago, to bring advanced ML techniques out of the lab (lots of active learning, for instance), and apply them at scale, for large enterprises where we can measurably prove impact. We're profitable, with multiple large scale Fortune 500 customers in pilots & in production.
There are 8 of us on the team, almost entirely technical – a mix of Machine Learning Engineers and Scientists with deep academic background and experience. The team includes 3 former founders. Co-founder and CEO founded Kimono Labs (YC W14, acquired by Palantir).
Competitive compensation, with material equity ownership for the right fit.
—
Apply by sending us an email: [email protected] | https://arena-ai.com
Interesting, thanks for putting this together. Curious – have you found good sources of granular data for credit card spending and movement patterns, perhaps from traffic/ride-sharing/mobile (that's anonymous/protects user privacy ofc)?
Today we are focused on pricing. We built a simulator that models hyper-specific market conditions and competitive dynamics to allow human users and AI agents to predict what will happen if they change the price of a product.
Pricing is just the start. We plan to create simulated, virtual copies of many more real environments - with the ultimate purpose of training a broader set of reinforcement learning agents for different tasks.
Today, we are focused on pricing. We built a simulator that models hyper-specific market conditions and competitor dynamics to allow human users and AI agents to predict what will happen if they change the price of a product.
Pricing is just the start. We plan to create simulated, virtual copies of many more parts of the physical world – with the ultimate purpose of training a broader set of reinforcement learning agents for different tasks.
Super interesting article. Curious, with the pharmaceutical developments we have already, what would be an upper bound for human AQ today? How might we measure it?
Yes, in fact we built API in partnership with EDGAR-online. It uses EDGAR-Online data but makes it available in a manner that's easier to digest and consume.