Topics
Latest
AI
Amazon
Image Credits:Tetsuwan Scientific
Apps
Biotech & Health
mood
Image Credits:Tetsuwan Scientific
Cloud Computing
Commerce
Crypto
Tetsuwan Scientific’s robotic AI scientist looks more like a glass cube.Image Credits:Tetsuwan Scientific
Enterprise
EVs
Fintech
fund raise
appliance
Gaming
Government & Policy
Hardware
Layoffs
Media & Entertainment
Meta
Microsoft
Privacy
Robotics
protection
Social
Space
Startups
TikTok
transportation system
speculation
More from TechCrunch
event
Startup Battlefield
StrictlyVC
Podcasts
video
Partner Content
TechCrunch Brand Studio
Crunchboard
Contact Us
Cristian Ponce was wearing an Indiana Jones costume when he meet his co - laminitis Théo Schäfer . It was at a Halloween company in 2023 thrown by Entrepreneur First , a inauguration program that introduces father to one another before they launch an mind .
The two hit it off , Ponce remembers . Schäfer had meditate at MIT with a master ’s in underwater independent robots and work at NASA ’s Jet Propulsion Lab search Jupiter ’s moons for foreign sprightliness . “ Crazy stuff , ” Ponce said , grin . “ I was coming from Cal Tech , doing bioengineering , ” where he work on E. coli .
The two bind over story about the grind of being a lab technician . Ponce ( see above leftfield ) especially complained about all the manual labor involved in familial engineering . The petty science laboratory tech can expend hours with a scientific syringe “ pipette , ” manually moving liquid from thermionic tube to tube .
attempt to automatise the process have not taken off because the robots capable of doing it are specialized , expensive , and ask exceptional programming attainment . Every time the scientist need to change an experimentation ’s parameters — which is all the sentence — they ’d have to wait for the software engineer to program the bot , debug it , and so on . In most cases , it ’s easier , cheaper , and more precise to practice a human .
The company they founded , Tetsuwan Scientific , limit out to treat this problem by modifying lower - cost whitened label lab robots .
But then in May 2024 , the co - founders were see OpenAI ’s multi - model product launching ( the one thatticked off Scarlett Johanssonwith a sound - alike voice ) . OpenAI showed people talk to the model .
It was the miss link Tetsuwan Scientific take . “ We ’re look at like this demented breakneck progress of prominent spoken communication modelling right before our eyes , their scientific reasoning potentiality , ” Ponce said .
Join us at TechCrunch Sessions: AI
Exhibit at TechCrunch Sessions: AI
After the demonstration , Ponce fire up GPT-4 and showed it an image of a DNA colloidal gel . Not only did the manakin successfully interpret what the paradigm was , but it in reality also identified a problem — an unintended desoxyribonucleic acid fragment lie with as a “ primer dimer . ” It then offered a very detailed scientific suggestion on what caused it and how to alter the conditions to prevent it .
It was a “ faint incandescent lamp moment , ” Ponce described , where LLM models were already capable of diagnosing scientific output but had “ no physical bureau to actually perform the suggestions that they ’re making . ”
The co - founders were not alone in exploring AI ’s usance in scientific discovery . Robotic AI scientists can be trace back to1999 with Ross King ’s robots “ Adam ” and “ Eve,”butreally kickedoffwith a seriesofacademic papersstarting in 2023 .
But the trouble , Tetsuwan ’s research showed , was that no computer software existed that “ interpret ” scientific intent — what the experimentation is look for — into machinelike slaying . For instance , the robot has no way to understand the physical qualities of the liquids it is pipetting .
“ That automaton does n’t have the context to have it off . peradventure it ’s a viscous liquid . Maybe it … is going to crystallize . So we have to recount it , ” he said . Audio LLMs , with hallucinations tamped down by RAG , can work with thing “ that are hard to code . ”
Tetsuwan Scientific ’s robots are not humanoid . As the photo shows , they are a square glass construction . But they are being progress to evaluate issue and make modifications on their own , just like a human would do . This necessitate building software system and sensors so the robots can empathise thing like calibration , melted class characterization , and other properties .
Ponce ’s eye light up when he talks about the ultimate name and address of this work : sovereign AI scientist that can be used to automate the whole scientific method , from hypothesis through repeatable termination .
“ It is the craziest affair that we could perhaps solve on . Any technology that automates the scientific method , it is the catalyst to inflated growth , ” he says .
He ’s not the only one to conceive this direction . Others working on AI scientists include non - profit orgFutureHouseand Seattle - basedPotato AI .
