In ads, influencer partnerships, and blog posts, Anthropic has continued to position itself as the intellectual, even creative, AI platform.
“There’s hope in hard questions,” explicates one ad. “Keep thinking with Claude,” suggests another. Anthropic’s models, notably, are named after literary forms, while its CEO, Dario Amodei, has styled himself as a thought leader through a series of heady blog posts.
The company’s message is that AI should enhance human cognition rather than offload it.
Thinking Machines, the AI lab founded last year by former OpenAI leader Mira Murati, is now putting its own spin on that approach.
Earlier this week, the startup announced the release of its first AI model, Inkling, which it says can be fine-tuned via its Tinker platform (released last fall). The company says Inkling has 975 billion parameters, including 41 billion active parameters. It’s also an open-weights model, meaning that core aspects of the model are publicly available.
Thinking Machines concedes that Inkling is not the most powerful AI model on the market. Instead, the company aims to compete on customization. It also emphasizes that its goal is to enhance human intelligence rather than replace it.
The approach echoes some of Anthropic’s language and appears aimed at a more intellectual vision of AI.
“For artificial intelligence to benefit from distributed knowledge, it must itself be distributed. Every organization is powered by the expert knowledge of its people, gained and expressed through their work,” a recent company blog said.
“We believe in AI that helps the organization cultivate that unique knowledge, not AI that extracts a snapshot of it and replaces it with a standard offering,” the blog continued. “This cultivation is an ongoing process that requires AI to work with people, not in their stead.”
With Inkling, Thinking Machines says consumers and enterprises can shape the AI in their own image rather than use a system that reflects the average preferences of the broader population. Inkling’s terminal even includes an option to instruct the model to communicate and organize its responses according to the user’s preferences, such as writing in the style of a New Yorker article or communicating only in riddles.
The model also communicates extensively about how it’s processing prompts. Users can control how much reasoning effort and token capacity Inkling expends on a task.
All of this suggests that AI companies don’t believe they’ll get far by selling consumers technology that replaces their thinking. Consumers appear to prefer something closer to an intellectual superpower.