Records Archive

Good evening. My name is Cristian Ponce, and I lead Tetsuwan Scientific, the organizer of this event. Thank you for joining us tonight at this beautiful venue. Thank you as well to our community partners, Tech Bio Transformers & Bay Area Lab Automators, specifically Vega Shah and Luis Villa. Studio 45's Michael Raspuzzi also provided invaluable advice for this event. These organizations have done incredible work in building the community that many of us have come to know and lean on. Thank you.

I would also like to thank our speakers, Dr. Will Serber from Ginkgo Bioworks, Dr. Jimmy Sastra from Monomer Bio, Mark Bissell from Goodfire, and my colleague & cofounder, Alex Rolfness.

The work of these groups represents the frontier of the rapidly evolving field which we know as autonomous science. Its north star is to achieve the automation of both the hands and the mind of the scientist, in order to automate the mechanisms of the scientific method in full. This field has witnessed exciting developments over the past 24 months, as interest has developed at breakneck speed.

Ginkgo partnered with OpenAI to set up GPT on its RAC automation platform and set it loose on optimizing cell-free protein expression. It achieved state-of-the-art results. Monomer built a fully automated organoid workcell that runs hands-free at ten times the throughput. Goodfire interrogated the genome-language model, Evo2, with sparse autoencoders, to discover that the model had taught itself biology. Exon boundaries, transcription factor sites, protein structure, all learned from a set of 9 trillion base-pairs without labels.

We must maintain a respect for the incredible complexity and chaos of biology. It does not represent a simple set of problems. There remains a long way to go in the automation of the knowledge work involved in research. But of course, the potential is incredible. After decades of groundwork laid across artificial intelligence and lab automation, there is no better time in history to be working at the intersection as there is now.

At Tetsuwan, our concern is the physical automation of science and the centralization of R&D infrastructure that follows it. Many unfamiliar with lab automation are often surprised to learn that this remains an open challenge. Some of you arrived in a self-driving taxi and some of you arrived from the bench where you were pipetting by hand. How could this be? Why have we not automated more of our experimentation?

As Abhishaike Mahajan writes, “Most lab protocols can be automated, they just often aren’t worth automating”. What does this mean? This means that we have the hardware capable of automating many experiments, but it seldom makes sense to invest the time and energy in porting experiments from the bench to the workcell. And so, lab automation has remained confined to the niche of assembly line work of high-volume, low variety experimentation. At its core, this is an interface problem. The end users of automated tooling, scientists, are unable to interface with the tools without automation engineers. Automation engineering is an incredibly complex discipline tasked with converting implicit protocols into explicit instructions under a set of constraints the scientist does not fundamentally understand. Where a scientist would see moving 100 mics of buffer to 100 mics of cells, an automation engineer sees the tip type, the aspiration speed, the blowout volume, the multidispending logic, the geometry of the meniscus of the well, and more.

This is not a new type of problem. Our colleagues in computing are well-familiar with it. There is no Oracle without SQL, and no AWS without EC2. Interfaces predate infrastructure, and it is our hope at Tetsuwan to develop both the necessary interface to automation, and the centralized infrastructure, often referred to as the cloud lab, that follows. This achievement would mark a turn towards experimentation as computation. The ability to articulate experiments in a common language on a shared platform will improve reproducibility, accessibility, and the speed at which we discover. Alex will share more about our work briefly.

Efforts to further automate experimentation see a lot of skepticism, because this space has seen a lot of death and failure. So what is different this time? I could tell you that never before in history has lab automation benefited from the capital and talent that it does today. I could tell you that humanity has never had access to the computational resources that it does today. Those would be factually correct answers, but in truth, this time is different because it must be different.

Over the past 14 months, we have witnessed the rapid and systematic deconstruction of many of our nation’s scientific institutions.

Men and women who came to this country in pursuit of new knowledge have been turned away. Once hopeful PhD students have been trapped in the purgatory of endless rotations, as they struggle to find a lab that can afford them. Hack and slash men have cut billions in federal funding for research, and blinded by ignorance have cancelled grants for topics such as transgenic mice. Since its second term, the current administration has overseen 1.4B in cuts from NSF grants, 9.5B in cuts from NIH grants, and cumulatively fired thousands of employees from these organizations.

Meanwhile, in the world of pharmaceuticals, with the exception of Lilly and Novo, many of the giants have moved away from the “risk” associated with R&D, increasingly offloading the derisking of early stage drug programs to small firms over the past 20 years. Several of these large “pharma companies” now resemble glorified private equity shops more than they do medicine makers. Today, small, emerging biopharma companies account for 70% of the clinical-stage industry pipeline. Yet, following 7 interest rate hikes in 2022, these small firms have struggled to find the funding needed from private investors, and following Trump's second term, have struggled to find the funding needed from public institutions. This class of small companies that keep us moving forward have at best been neglected, and at worst, have been exploited.

In academia, a perverse incentive structure rewards novelty over rigor. Findings that can't be replicated flood the literature because it is not in the interest of highly lucrative journals to publish null results. Principal investigators are made to spend too much time chasing funding, leaving less time for research. Graduate students and postdocs, the labor force of academic research, are paid survival-level salaries for years with vanishingly small odds of a tenured faculty position waiting for them at the end.

The scientist in America, by no means, has failed. Yet, they themselves have been failed by the systems meant to serve them.

I don’t know what will happen next. Not for us, not for this country, and not for the process of scientific discovery. None of us do. Many of my colleagues, as I imagine many of yours, have come to believe, that the end of a great era of science in America is but a foregone conclusion.

But I do not share this belief. While the scientist has been challenged to make do with less, the toolmaker has continuously augmented the capital, time, and labor that we do have, to keep discovery moving forward. If we know we are going to have less resources, we must develop tools to make more out of less. This is the burden of the toolmaker: they must innovate so that we all can continue to innovate. There has never been a time more urgent than now for the development of new tools for research, because there has never been a time more urgent than now to make do with less. Whether these tools are molecular, mechanical, or digital, we need more and we need better.

This burden is intimidating, but we must remember that in the life sciences, we come from a long lineage of great toolmakers. It was a humble cloth merchant who crafted the lens that first witnessed microbial life. It was a chemist at Cetus with extra time on his hands following the automation of his synthesis work, who in myth, rode the double helix to invent the polymerase chain reaction. The cost of sequencing, described empirically by the Carlson Curve, has fallen faster than the cost of compute, as described by Moore’s Law, thanks to generations of ingenious engineering, from Sanger’s chain-termination chemistry to Ultima’s silicon wafers. What once took 2.7B dollars and 23 years can now be done for just a hundred dollars and within a day.

Our great tradition of toolmaking is continued by many in attendance today, from our friends at Bay Area Lab Automators, to those in the open source community like Rick Wierenga with PyLabRobot. And of course, a class of promising startups in the crowd, including Zeon, Genetic Assemblies, Olden Labs, OnePot.AI, and of course, Ginkgo, Goodfire, and Monomer. At Tetsuwan, we are honored to work on these problems alongside you all. There is no greater mission that we may be called to than to serve our scientists by improving the tools that shape our understanding of the natural world.

Of course, the toolmaker has never been presented with a set of tasks so large or consequential as the set they are presented today. They are challenged to develop new disease models with stronger predictive validity, expand the domain of automation to build centralized research infrastructure, and develop tools capable of recognizing patterns invisible to the human mind.

Yet in the pressure we feel, we must also remain sober-minded about the consequences we may bring about. To truly believe that the automation of the scientific method is possible and lies on the horizon, is to truly believe that our most consequential invention as a species lies on the horizon. If the dreams this field sells become a reality, are we prepared? How would exponential scientific progress impact society? How do we ensure the fruits of discovery are shared amongst us all, and not privatized by a select few?

We are faced with many challenges and many questions. May we rise to them as a community, and may we remember our place: as toolmakers. We should not aspire to substitute or otherwise replace the scientist, it is our job to serve the scientist.

I’d like to end these opening remarks with a quote from my favorite speech written about biology, delivered by Robert Sinheimer in 1967 at Caltech’s Beckman Auditorium Hall, or as many students called it, the wedding cake.

“Ours is an age of transition. After two billion years, this is the end of the beginning. It would seem clear, to some achingly clear, that the world, the society, and the man of the future will be far different from that we know. Man is becoming free, not only from the external tyrannies and caprice of toil and famine and disease, but from the very internal constraints of our animal inheritance, our physical frailties, our emotional anachronisms, our intellectual limits. We must hope for the responsibility and the wisdom and the nobility of spirit to match this ultimate freedom. We must ask that the changes we introduce be orderly and with humanity aforethought. At Caltech, and in all of science we have been, in a sense, children, spewing change into society with scant thought for the consequence. We in science are growing up now. Our toys become more potent. The little games we play with nature are for great stakes, and their outcome moves the whole social structure. We must accept our responsibility.”

Thank you.