The startup Inflection AI, launched by LinkedIn co-founder Reid Hoffman and DeepMind co-founder Mustafa Suleyman, debuted a powerful new Large Language Model (LLM) on Thursday.
The company, which has raised $225 million to develop “personal AI for everyone,” says the model, called Inflection-1, will run thousands of NVIDIA machines on “extremely large datasets.” We mentioned that it was trained using a graphics processing unit (GPU).
The technical note also notes that Inflection-1 outperforms many similar products on common benchmarks used to compare LLMs, including GPT-3.5, an OpenAI model that powers ChatGPT, and LLaMA, created by Meta. claims to be
Inflection in May announced Pi, a personal AI model designed specifically for conversing with humans, rather than answering questions and improving productivity.
“The Pi is a new breed of AI that is not only clever but has great EQ,” Suleiman, who is also CEO of Inflection, said in a statement.
Inflection-1 powers Pi, but the chatbot was trained beyond the models the company releases through its API.
Tech giants and start-ups alike are racing to release or improve ChatGPT alternatives to capture a slice of the growing generative AI business. AI startup Anthropic has Claude and Stability AI has StableLM. Google has Bard, and Amazon announced in April that it was improving LLM, which powers the company’s voice assistant Alexa.
