Contextual AI emerges from stealth with $20M to pursue ‘artificial specialized intelligence’

Contextual AI, a brand new enterprise-focused AI startup based in Palo Alto by veterans of Fb AI Analysis (FAIR) and Hugging Face, immediately emerged from stealth and introduced it has raised $20 million in a seed spherical led by Bain Capital Ventures (BCV), with participation from Lightspeed, Greycroft and an extended checklist of different angel buyers.

The corporate has a singular focus: Creating what its co-founders Douwe Kiela and Amanpreet Singh refer to in a blog post as “‘ASI’—synthetic specialised intelligence.”

“Why waste parameters, cash, latency and compute on a mannequin that is aware of Shakespeare and quantum physics, when all it ought to actually do is remedy your organization’s drawback?” they write, later taking a public shot at the most well-liked and extensively used LLM up to now, OpenAI‘s ChatGPT.

The chance Contextual AI sees

The duo has current high-profile expertise within the area of AI and huge language fashions (LLMs). Each beforehand labored at Meta Platforms’ FAIR on growing methods for detecting hate speech in memes and the sale of illicit items on Fb. At HuggingFace, they labored on LLMs.

Now, with the launch of Contextual AI, they search to use their abilities — alongside different engineers (the entire crew measurement is 12 thus far) — to resolve six issues they are saying confront enterprises in search of to make use of present LLMs.

The issues Contextual AI is concentrating on embrace:

• Knowledge privateness
• Customizability
• Hallucinations
• Compliance
• Staleness, which the Contextual founders describe of their weblog submit because the comparatively outdated data contained in main AI fashions: “ChatGPT doesn’t know something that occurred after September 2021, and probably the most downloaded mannequin on the HuggingFace hub is BERT, which nonetheless thinks that Obama is president.”
• Latency and value

Committing to open supply AI

The founders additionally took their funding announcement to pledge a dedication to the burgeoning open supply AI group, which has come beneath renewed scrutiny within the final week by U.S. lawmakers (particularly concentrating on Meta Platforms’ “leak” of its LLM LLaMA.)

“With a expertise as vital and disruptive as LLMs, we imagine there must be extra gamers, and that it is vitally harmful to have this expertise solely within the fingers of some incumbent massive tech firms,” the wrote of their weblog submit. “Our method each helps and is supported by open supply. We plan to leverage open supply software program as a lot as we are able to and to offer again to the group in return.”

For its half, BCV stated they supported Contextual AI’s targets and ambitions and noticed monumental potential available in the market for enterprise-grade AI.

“Our view is that AI will turn into ingrained in all the things enterprises do in methods which can be safe, deeply built-in and most significantly, contextual to worker workflows,” BCV principal Rak Garg and associate Aaref Hilaly wrote in a blog post. “Spending time with Douwe and Aman, it’s clear they each perceive that at an intuitive degree. We’re thrilled to be working with them each to carry the ability of language fashions into the office.”