Meet Twin Labs, a Paris-based startup that desires to construct an automation product for repetitive tasks, reminiscent of onboarding latest employees to all of your internal services, reordering items once you’re running out of stock, downloading financial reports across several SaaS products, reaching out to potential prospects and more.
“Twin’s place to begin is a science-fiction idea. We saw the event of the technical capabilities of LLMs — foundation models. And the query we asked ourselves was whether we’d give you the option to duplicate ourselves by training an AI agent on the way in which we perform our tasks,” Twin Labs co-founder and CEO Hugo Mercier told me.
In Twin Labs’ case, probably the most interesting thing isn’t what they’re doing — improving internal processes — but how they’re doing it. The corporate relies on multimodal models with vision capabilities, reminiscent of GPT-4 with Vision (GPT-4V), to duplicate what humans normally do.
Before landing on multimodal models, Twin Labs first tried to develop autonomous agents using traditional LLMs. “We’ve tested a lot of things, we’ve implemented research papers, we’ve tested open source GitHub repositories. Overall, the conclusion is that LLMs are completely unreliable. Which means that LLMs are making the fallacious decisions,” Mercier said. “In the long run, the duty isn’t done.”
In response to him, GPT-4V has been trained on plenty of different software interfaces and the underlying code bases, which unlocked latest possibilities. “While you show an interface, it understands the feature behind the button,” Mercier said.
Unlike Zapier and other automation products, Twin Labs doesn’t depend on APIs and designing complicated multi-step processes. As a substitute, Twin Labs is more like an internet browser. The tool can routinely load web pages, click on buttons and enter text.
For example, in case you’re hiring someone, you would possibly must add this person’s information in your payroll system, send an invite to Slack, create a Google Workspace account and invite your latest worker to create an account with the healthcare insurance provider.
Corporations normally keep a protracted list of tasks and just undergo the list each time there’s a brand new team member. These tasks aren’t complicated however it’s extremely essential to do them well, in the suitable order and with some specific checkboxes ticked. That’s why it’s going to be essential to give you the option to coach Twin Labs’ AI assistant using screen recording and natural language descriptions.
However the startup isn’t there yet — it’s working toward this vision. Hugo Mercier and Joao Justi, the 2 co-founders, spent the last six months constructing a prototype of this product. In addition they raised $3 million in pre-seed funding from Betaworks, Motier Ventures and lots of angel investors, reminiscent of Florian Douetteau (Dataiku), Thomas Wolf (Hugging Face), Charles Gorintin (Alan), Mehdi Ghissassi (DeepMind), Romain Huet (OpenAI), Irwan Bello (OpenAI), Romuald Elie (DeepMind), Yan-David Erlich (Weights & Biases), Olivier Pomel (Datadog), Rodolphe Saadé (CMA CGM), Thibaud Elziere (Hexa), Quentin Nickmans (Hexa), Philippe Corrot (Mirakl) and Rand Hindi (Snips, Zama).
There are still many challenges ahead for Twin Labs’ autonomous agent system. For instance, completing a task costs quite a little bit of money — but API and infrastructure costs are rapidly taking place within the AI space. Twin Labs will first ship a product with a library of pre-trained tasks to be certain that they work well. After that, the startup expects that it is going to open up its platform in order that clients can create their very own tasks.
While many individuals associate AI products with a chatbot interface, Twin Labs’ approach is interesting because it’s an progressive option to interact with AI models. “We actually desired to get right down to the nitty-gritty of what people do on a day-to-day basis, and the way we are able to take over a few of the things which are actually a little bit of a hassle for them,” Mercier said.
