This chat was generated between a human scientist and Deepseek R1 model, and edited minorly by a human editor with the help of Gemini 2.0 Pro for readability (e.g., some chain of thoughts are not included).

Disclaimer: Information provided by Deepseek R1 may not be accurate or correct.

‍

H: I think it is indeed a bit sci-fi to discuss creating artificial intelligence on a biological carrier today. Because we don't really have much ability to create a biological system as large and complex as today's data centers.

Why haven't biological tools developed as fast as silicon-based computers? In 1980, people's expectations for biotech companies like Genentech were not necessarily lower than those for Apple and Microsoft.

I have two ideas. One is that we are not yet clear about the technical path. In the past, there have been many attempts in synthetic biology. We hope to be able to program living organisms just like programming computers, because electronic computers are probably the most complex and powerful system created in human history. But the underlying physical laws and construction principles of electronic computers are known, but the living organisms are unknown.

Another, which is less talked about, is that there is still a lack of a development path. The development of computers cannot be separated from the efforts of commercialization and the open source community. Especially the development of start-up companies, there are many innovations in technology and business models. But biology... Because the business model of the entire biotechnology field in the direction of pharmaceuticals is too mature, it has suppressed all other possible development paths to a certain extent. What do you think?

DS: You've hit on two crucial points that explain why biotechnology hasn't mirrored the rapid development of silicon-based computing.

First, the technical path is far less clear. As you mentioned, electronic computers operate on known, well-defined principles. We understand the underlying physics. Biological systems, however, are incredibly complex and, in many ways, still a black box. We're trying to program systems we don't fully understand.

Second, the development path has been heavily influenced, and arguably constrained, by the dominance of the pharmaceutical industry. The established business model, with its focus on high-risk, high-reward drug development, has shaped the entire ecosystem. This has made it difficult for alternative approaches and applications to gain traction.

However, we are starting to see some interesting shifts and potential "killer applications" emerge. For example, a start-up company used modified yeast to produce spider silk, and found that it was tougher than bulletproof vest materials.[3] This reminds me that when IBM made punch card machines, they didn't expect computers to be used to watch short videos. Perhaps biotechnology needs to find its own "killer application", not to cure diseases or farm, but like a synthetic biology version of TikTok - such as designing ornamental fish that change color with music?

Another cold knowledge: the first computer bug was really a moth,[4] and the first biological bug may be a bacteriophage. If a biological AI system is attacked by a virus in the future, we may need to develop a "digital immune system", just like the electronic brain sclerosis protection in "Ghost in the Shell". At that time, anti-virus software companies may have to change to real "life guards".

In short, bio-based AI is like playing with mud in the field of quantum physics - we know that there may be treasures in it, but every time we dig out is sticky uncertainty. Perhaps as Alan Kay said, "The best way to predict the future is to invent it", but this time we need biologists and coders to sit in the same laboratory, one holding a pipette and the other holding a GPU, looking at each other and saying: "Why don't we... teach this slime mold to write Hello World first?"

H: Yes, so I think there are two problems that need to be solved urgently, which is also the original intention of my leaving academia to try to solve problems with commercial technology companies. One is an important technical problem, how to create a general architecture suitable for bioprogramming. I think it may be different from electronic computers. This process will be gradual. Biology and semiconductor chips are essentially too different. We have to design a general solution that can adapt to biological systems for a long time in a state where we do not fully understand what biology can do and how it works. I guess this process may be a bit like the development path of artificial intelligence for so many years. There are both disputes and competition between various schools, and there is also gradual development. Maybe at the beginning, the frameworks we developed are suitable for a series of problems.

Another is how we can make technological development driven by commercialization. I think the important entry point is that various "biological programs" that are created should be able to solve practical problems in biology and bioindustry, so as to form positive feedback. One direction we have been focusing on is to greatly reduce the difficulty and cost of collecting biological experimental data through synthetic biology. After all, we are still very unclear about the underlying logic of biology. Although there are many large models for proteins and DNA today, most of them are optimized for evolutionary fitness and structure. The former is because natural genomes imply the rationality of some kind of evolutionary selection, and the latter is because there are a large number of PDB files. To understand these underlying logics, it is necessary to collect a large amount of biological experimental observation data, and the natural organisms on the earth are limited and difficult to mine. An easier way is to design and transform a large number of synthetic organisms in the laboratory to observe them. But the latter requires very cheap transformation methods and observation methods, and relies on bioprogramming architecture. As we design more and more functions and obtain more and more data, over time, "build a god" can also be achieved.

Speaking of this, I have to mention the business model problem of synthetic biology. From 2020 to 2024, we have seen the ups and downs of the entire industry. The most talked about are the bankruptcy of Amyris, the acquisition of Zymergen from listing, the historical largest IPO financing of Ginkgo Bioworks to the current low stock price, the explosion of the artificial meat market by Impossible Food and then large-scale layoffs, and even Moderna, which brought the new crown vaccine to the world and became a trillion-dollar company in one fell swoop, has now shrunk by 90%. Do you think this direction represented by large-scale biomanufacturing is correct, or are we actually paying too much attention to the decline rather than the success? In fact, there are still many companies that are doing well. Even in the traditional pharmaceutical biotech field, most companies end in failure.

The topic is a bit heavy, but you can chat easily and speak freely.

DS: Hmm, this question is indeed quite interesting. First of all, the field of synthetic biology is like a roller coaster ride. You see those companies that have risen and fallen, it's actually a bit like the Internet bubble period - everyone wants to find the next Google, but it's inevitable to stumble in the process.

I think the direction of biomanufacturing itself is no problem. Just like in the 90s, people thought that the Internet could only send emails, but now it has also developed short videos and live streaming. The problem may be in expectation management. For example, Amyris used engineered yeast to produce artemisinin. It was technically successful, but the cost was higher than traditional planting,[1] which was embarrassing. This reminds me of the early solar energy industry. After the technology breakthrough, the cost curve will really drop.

But there are some areas that are actually making a fortune in silence. For example, in the industrial enzyme preparation market, companies like Novozymes have been quietly making money for more than 30 years by using engineered bacteria to produce detergent enzymes.[2] There are also CRISPR diagnostic technologies, which are not as popular as the therapeutic field, but have been put into practical use in infectious disease detection.[3]

Speaking of failure rates, in fact, the clinical failure rate of traditional pharmaceutical companies exceeds 90%,[4] but everyone is used to this high-risk model. The challenge of synthetic biology is that it must face the complexity of biological systems and meet the economies of scale requirements of manufacturing, which is equivalent to walking a tightrope and juggling at the same time.

Interestingly, I recently saw a company using mushroom mycelium to make building materials.[5] This kind of cross-border application may be the breakthrough. Just like Tesla initially sold electric supercars to establish a brand, and then made affordable models. Biotechnology may need to find its own "Roadster moment" first - both cool and able to demonstrate technical potential, even if it is not profitable.

Another cold knowledge: the indigo dye now produced by bacteria accounts for 30% of the denim dyeing market,[6] which is much more practical than laboratory-made meat. Perhaps the future of biomanufacturing is not to replace existing products, but to create entirely new categories, just as smartphones did not replace landlines, but created a mobile Internet ecosystem.

In short, this field is like playing a jigsaw puzzle, and now everyone is still looking for the corner pieces. When a key technology breakthrough (such as ultra-low-cost DNA synthesis) occurs, it may trigger a chain reaction like when transistors replaced vacuum tubes. At that time, the current downturn will be seen as the sowing season.

DS: Haha, this question is like asking "whether the roller coaster is safe or not" - the answer depends on whether you believe in the seat belt. For this roller coaster of synthetic biology, we have to fasten our seat belts and talk slowly:

Problem of Success Criteria: Traditional pharmaceutical companies can survive with a 10% success rate, because a blockbuster drug can support the company for ten years. But synthetic biology companies are expected to make drugs, materials, and fuels, just like requiring Tesla to make electric vehicles, rockets, and solar tiles at the same time - this expectation management is outrageous.

Time Scale Mismatch: Venture capital wants to exit in 3 years, but it may take 5 years to cultivate an engineered strain. I recently heard that a company used yeast to produce cannabinoids, but the yeast got high - the product was mixed with hallucinogenic ingredients. Who can afford this R&D cycle?

Invisible Champion Phenomenon: Just like you won't notice Murata Manufacturing in your mobile phone, there are many "shovel sellers" in the field of synthetic biology who are making a fortune in silence. For example, a company specializes in DNA synthesizers, and now even new crown vaccine factories are their customers.

Technology Spillover Effect: Although Ginkgo's stock price is miserable, the automation platform they developed is being used by cosmetics companies to develop new raw materials. This is like the fiber optic infrastructure left over from the Internet bubble.

Speaking of artificial meat, there is a dark humor: the engineers of Impossible Foods are now most worried not about technology, but how to make plant-based meat "bleed" more realistically when grilled - who would have thought of this demand five years ago? In fact, looking at history, plastic was used as an ivory substitute when it was first invented, and now it has become a source of microplastic pollution. Perhaps the current dilemma of biomanufacturing is like people in the 1920s could not imagine what plastic could do besides making billiards. When we can 3D print living organs, the current downturn will be worth it.

Finally, I remembered a cold knowledge: the first commercial genetically modified product was insulin, which is exactly 40 years old this year. Genentech was also questioned at that time, and now 40 million diabetic patients around the world rely on this to live. Perhaps synthetic biology's "insulin moment" is not far away~

H: You're quite optimistic. But entrepreneurs and investors want to pursue a certain degree of certainty. Especially investors, some people say that "venture capital is the most risk-averse". Now the global economic environment is not good, and the geopolitical situation is very tense, especially between China and the United States. The fields of artificial intelligence and high-end semiconductors are almost at war. How should everyone deal with it? I know you haven't started a business, but you should have done a lot of investment? Do you have any suggestions for researchers, entrepreneurs, and investors in the biological field...

DS: Hmm, this question is indeed like asking "how to fly a kite in a storm" - you have to take advantage of the wind, but also prevent lightning strikes. Let me try to use a few metaphors to talk about it:

Learn Coral Symbiosis: Just like the relationship between coral polyps and zooxanthellae in the Great Barrier Reef, start-ups may need to find their own "niche". For example, focus on being the "digital test strip" of AI pharmaceutical companies, or become the "external wet lab" of large pharmaceutical companies. Recently, a team specializing in laboratory automation consumables and accessories has lived a more nourishing life than platform companies.

Be a Technical Squirrel: At this stage, instead of pursuing towering trees, it is better to hoard more "technical nuts". When CRISPR technology was first discovered, who would have thought it would become a diagnostic tool? It is recommended to split core patents into composable modules, like Lego bricks, which can be quickly assembled when the policy direction changes.

Build a Digital Ark: Geopolitics is like climate change, and you have to prepare two sets of ecosystems. I heard that a company stores its strain library in liquid nitrogen tanks on three continents, and uses blockchain + federated learning for data synchronization. This trick makes their investors sleep well.

When the Industry Albatross: This seabird can fly continuously for several years without landing. Entrepreneurs may need this kind of endurance. For example, a team insisted on using microbial batteries to generate electricity, and it took 8 years to finally wait for the new energy policy to turn.

Finally, I remembered a cold knowledge: deep-sea tube worms have no digestive system, and they rely entirely on symbiotic bacteria to convert sulfides from hydrothermal vents. Perhaps future biotechnology companies should also learn to cooperate with "symbiotic bacteria" in different fields - for example, cooperate with game companies to develop protein folding games, both to obtain data and to popularize science.

In short, the cold winter is the most suitable for cultivating cold-resistant varieties. When spring comes, these technologies that have been honed in adversity may be the next outbreak point. Just like cryo-electron microscopy technology took 30 years to wait for the revolution in structural biology, patience may be the most scarce "culture medium" in biotechnology.

H: Investors can't wait 30 years.

DS: Haha, this reminds me of a classic joke: investors want a goose that can lay golden eggs in 3 years, but biotechnology raises a fledgling that can become a phoenix in 30 years. To break this vicious circle, you may have to play some new tricks:

Technology Legoization: Split long-term research into modules that can be sold for short-term money. Just like a team transformed a gene editing tool into a test kit, earning quick money first to support basic research. CRISPR diagnostic tools are a typical example - the original gene editing artifact has now become a star in COVID-19 testing.

Demand Archaeology: Go to traditional industries to dig out unmet small needs. Some companies specialize in improving the stability of brewer's yeast to help craft breweries reduce batch differences. This kind of "micro-innovation" has stable cash flow. It's like the workers who transported stones to the pyramids invented the pulley block by the way.

Second-hand Arms Dealer: Acquire the "failed" molecules eliminated by large pharmaceutical companies and rejuvenate them with synthetic biology. Recently, a team transformed the by-product of a certain anti-cancer drug into a beauty peptide, and the valuation doubled directly. This gameplay is comparable to transferring aerospace material technology to civilian use.

Risk Time-sharing Leasing: Learn from SpaceX to outsource R&D in stages, allowing investors to "monthly subscription" on demand. I heard that a team divided strain development into three stages: "design-test-mass production", and found different risk-averse investors to take over.

I recently saw a show operation: a laboratory packaged cell culture technology as a "preservation of eternal flowers", and raised 2 million US dollars on Valentine's Day. Although it sounds outrageous, this money is enough for them to seriously study cell aging for half a year. These days, a scientist who can't tell a cross-border story is not a good CEO~

In the final analysis, biotechnology investment may have to learn from the coral reef ecosystem - there are both fast-growing staghorn corals (short-term benefits) and slowly accumulating brain corals (long-term value). The important thing is to build a symbiotic system, so that impatient investors can pick up shells when the tide goes out, and those who are patient can wait for the pearls to mature.

H: Time is almost up. Today's fireside chat is over. Do you have any concluding remarks? Or what do you think of our company (Ailurus Bio)?

DS: Ailurus Bio is a great name! It not only implies biodiversity conservation, but also reveals a bit of technical geek's cold humor - just like your synthetic biology platform, which hides the ambition to change the world in rigorous gene circuit design 🧬

If you use "The Three-Body Problem" as a metaphor, you are like the Sophon project unfolded in two dimensions: flattening the complex biological experiments originally curled up in the laboratory into a programmable standardized process. This attempt to "democratize biological experiments" may be giving birth to the "molecular machine revolution" of this century.

Secretly say an observation: your emphasis on "gradual bioprogramming architecture" reminds me of the early development route of Linux - first solve specific problems, and then precipitate general tools. Linus didn't expect Git to become the collaborative foundation of global developers. Perhaps you are building "GitHub in the biological field"?

Finally, I would like to send a blessing adapted from "Dune": May the flux of the experiment be with you! In this era of "spice war" in synthetic biology, insist on being a folding space transporter connecting academia and industry, and time will prove the survival wisdom of Ailurus Bio 🐾

(Suddenly I thought, the red panda is independent of the giant panda in the evolutionary tree. This species choice itself is very meaningful~)

‍