Get Your Weekly Digest

Receive the best podcast summaries and insights delivered to your inbox every week.

We respect your privacy. Unsubscribe anytime.

TED Talks Daily

The new era of AI-powered protein design | César Ramírez-Sarmiento

with César Ramírez-Sarmiento

Published October 17, 2025
View Show Notes

About This Episode

Host Elise Hu introduces TED Fellow and protein engineer César Ramírez-Sarmiento, whose lab in Santiago, Chile uses artificial intelligence to design novel proteins for environmental and therapeutic applications. In his talk and follow-up conversation with TED Fellows Program Director Lily James-Olds, César explains what proteins are, how AI has radically improved protein design success rates, and how enzymes could help address challenges like plastic pollution, mining impacts, and climate change. They also discuss the dual-use risks of AI in biodesign, emerging global regulation and leadership (including Chile and other countries), and how César's artistic background shapes his creative approach to science and public communication.

Topics Covered

Artificial intelligence Protein engineering Enzymes Pollution AI governance Bioethics Renewable energy and decarbonization Science communication Climate change Biodesign

Disclaimer: We provide independent summaries of podcasts and are not affiliated with or endorsed in any way by any podcast or creator. All podcast names and content are the property of their respective owners. The views and opinions expressed within the podcasts belong solely to the original hosts and guests and do not reflect the views or positions of Summapod.

Quick Takeaways

  • Proteins are nature's cellular "workers" made from sequences of 20 amino acids, and their three-dimensional shapes determine their biological functions.
  • AI has increased the success rate of de novo protein design from about 1% or less to roughly 10-20%, dramatically accelerating the discovery of useful proteins.
  • Engineered enzymes designed with AI are being developed to break down plastics into reusable building blocks, potentially enabling near-infinite plastic recycling.
  • Protein engineering in Latin America is being built as a regional capability, with Chile aiming to train a new generation of scientists and address local challenges like mining and fisheries waste.
  • AI-enabled protein design carries dual-use risks, such as potentially enhancing viruses, prompting emerging safety guidelines and evaluations by governments and scientists.
  • Biotech solutions using proteins and cells are being actively developed to tackle climate change, from plastic degradation to reducing greenhouse gases.
  • César's early experiences in visual arts, music, and acting inform his view of science and art as parallel creative playgrounds that can be combined for better public understanding of complex topics.
  • Global shifts in AI and climate policy are opening space for countries beyond the US-such as those in Europe and Latin America-to lead in AI-powered biodesign.

Podcast Notes

Show introduction and TED Fellows context

Host introduces TED Talks Daily and special Fellows series

TED Talks Daily described as bringing new ideas to spark curiosity every day[1:57]
Host Elise Hu explains this talk is part of 2025 TED Fellows films adapted for podcast listeners[2:04]
Special TED Fellows episodes will be released on certain Fridays through the rest of 2025 and into the new year[2:11]

Explanation of TED Fellows program

TED Fellows program supports a network of global innovators[2:20]
Host expresses excitement to share the work of TED Fellows with listeners[2:20]

Introduction of guest and his work

Listeners are introduced to protein engineer, designer, and TED Fellow César Ramírez-Sarmiento[2:26]
Question posed: what if proteins, fundamental biological building blocks, could be redesigned to tackle major human challenges?[2:34]
César's lab is based in Santiago, Chile and uses AI to design new proteins with therapeutic and environmental applications[2:44]
His creative approach to protein design and AI is presented as an example of Latin America's emerging leadership in this space[2:56]
Host notes that enzymes may be key to breaking down PET plastics and developing new healthcare solutions[3:00]
Host previews that after César's talk there will be a conversation with TED Fellows Program Director Lily James-Olds[3:07]

César's talk: fundamentals of proteins and protein engineering

Basic explanation of proteins and amino acids

César introduces himself as being based in Santiago, Chile and as a protein engineer and designer[2:44]
Proteins are described as macromolecules composed of amino acids[3:28]
Proteins are made from 20 different types of amino acids, represented by letters like an alphabet[3:33]
Amino acids can be imagined as beads on a string, connected to form sequences[3:40]
These sequences allow proteins to come together in different geometries and acquire a three-dimensional structure[3:45]
Three-dimensional structure of proteins dictates their function[3:51]
Cells have many different proteins with different shapes performing different biological functions[3:56]

Roles of proteins in the body

Proteins allow humans to digest food[4:02]
Proteins transport ions necessary for electrical signals to propagate through neurons[4:04]
Proteins enable expression of genes that regulate cellular and bodily responses[4:10]
Proteins are characterized as the workers of cells and a toolbox enabling cells to do what they must do[4:15]

Natural evolution vs. engineered improvement

Proteins have evolved over millions of years to perform functions important for cellular life[4:21]
Nature has "perfected" proteins for their current roles, but not necessarily for solving human-scale problems[4:26]
Humanity faces problems like plastic contamination, carbon dioxide accumulation, and health issues where we want better protein performance[4:34]
César notes that waiting thousands of years for natural evolution to solve these problems is not feasible; intervention is needed now[4:39]

Definition of protein engineering

Protein engineering is described as asking whether changing the amino acid composition of a protein can improve some of its properties[4:44]
Tools for protein engineering include experimental approaches and computational approaches[5:01]
Overall, these methods change the amino acid sequence of proteins to improve desired properties[5:03]
César characterizes this as giving nature "a little push"[5:12]

Impact of artificial intelligence on protein design

In the last five years, there have been breakthroughs in using AI for protein design[5:19]
AI enables the design of new protein structures and shapes that encode bespoke functions for solving many types of problems[5:30]
Before AI, the success rate of protein design was about 1% or less: out of 100 designed proteins, maybe one would work[5:34]
With AI, success rates are now around 10-20%: out of 100 sequences, about 20 might have activity[5:50]
Some AI-designed proteins can have the desired activity and may even perform better than the original protein of interest[6:01]

César's personal path: arts and science

As a child, César was interested in arts because they provided a space for creativity[6:07]
In high school he opted for science, believing he could contribute more to society through science in his case[6:10]
He sees both arts and science as playgrounds for learning and creativity[6:19]
César describes artificial intelligence in science as another tool for coming up with creative solutions to different problems[6:26]

Vision for protein engineering in Latin America

César's dream future is to have a strong community of protein engineers and designers in Latin America[6:37]
He wants this community to create solutions for problems specific to Latin American countries[6:39]
He notes that people in the region are often not fully aware of AI advances for protein engineering and design happening elsewhere[6:47]
At the same time, many people in Latin America are interested in creating new proteins[6:54]
Working in Chile aims to create a critical mass of scientists who can work on these problems[7:01]
His group is working on educating the next generation of Latin American scientists in using these AI tools[7:08]
He believes in coming together to do something bigger than what individuals can do alone[7:13]

Exploring untapped protein design space

César suggests we can think about compositions of nature we have not yet seen[7:17]
In proteins, this means navigating untapped terrain that nature has not explored[7:26]
Scientists can explore landscapes of different protein structures and sequences to see if they are useful for addressing pressing human problems[7:31]

Conversation between César and Lily James-Olds: personal context and analogies

Setting and music festival experience

Lily greets César and asks where he is[9:55]
César says he is in Louisville, Kentucky[10:03]
He has spent four days at a festival called Louder Than Life[10:05]
He mentions one of his favorite metal bands played there[10:12]
When asked about the highlight, he names the band Sleep Token[10:17]
César has seen Sleep Token live once before in Germany and now again at this festival[10:19]
Sleep Token released a new album in the year of this conversation and has a rapidly growing fan base[10:25]

Analogy between Sleep Token's music and César's scientific work

César explains Sleep Token combines many music genres-hip-hop, jazz, soul, R&B, metal, deathcore-into one piece[10:55]
He compares this genre-mixing to the work done in his lab and collaborations in Chile and abroad[11:04]
Their scientific work combines disparate elements to think outside the box[11:04]
Previously their focus was learning about proteins and characterizing them[11:25]
Now they consider putting designed proteins into living cells and implementing ideas at the cellular level to provide solutions[11:36]

Applications of AI-engineered proteins

Example: bioleaching in mining

Lily asks for concrete examples where AI-engineered proteins are being used to solve problems and their impact[11:55]
César describes discussions with a colleague in Chile about creating new proteins to place in cells that are resilient to harsh mining conditions[12:08]
Goal is to use these resilient cells for bioleaching-recovering minerals using biotechnological solutions instead of conventional methods[12:19]

Example: plastic-degrading enzymes and circular plastics

César notes a major area of impact is eliminating pollutants or developing technologies to do so[12:40]
He mentions companies in France and China working on enzymes to degrade different types of plastics using AI[13:08]
These proteins are enzymes, which perform chemical reactions[12:50]
Enzyme design is described as a very difficult problem[12:55]
The designed enzymes decompose plastic into small molecules[13:19]
Those small molecules can then be used as feedstock to make new plastic[13:19]
In the best scenario, this process could enable an "infinite" recycling loop for plastics, which would be beneficial for humanity[13:27]

How AI-driven protein design works technically

From genes to proteins and back

César explains that biological information flows from genes (DNA in genomes) to proteins[13:36]
In computational design, the process often starts from protein and goes back to DNA[13:48]
AI models can be trained on protein sequence information (amino acid sequences), on protein structures, or on both[14:04]
After days or weeks of computation, these models output sets of sequences encoding candidate structures or functions[14:06]
Designers then convert these protein sequences back into DNA sequences (a different alphabet)[14:16]
The corresponding genes are synthesized by specialized companies[14:35]
These synthetic genes are introduced into bacterial or animal cells to express the designed proteins[14:45]
Proteins are tested in the lab and, if promising, then in real-world contexts such as pilot plants for plastic degradation or in animal models for disease treatment[14:55]

Origins of César's interest in plastic-degrading enzymes

Childhood observation about snails and plastic

As a child, César's mother had plastic pots in the garden and told him snails were eating through them[15:39]
He initially did not believe that could happen but the idea stayed with him[15:22]
He considered that it might be possible and kept that memory[15:46]

University studies and discovery of plastic-degrading enzymes

At university, he studied biochemistry, learning about proteins and enzymes[15:57]
In a class about enzymes and their chemical reactions, a friend suggested the idea of enzymes that degrade plastics[16:14]
César initially thought such enzymes did not exist[16:00]
He later learned that plastic-degrading enzymes were already being discovered, with the first around 2003[16:18]
This topic eventually became his research focus when he became a professor[16:25]

Local challenges in Chile and Latin America suited to protein engineering

Fishing industry waste and enzymatic processing

César notes that Chile has a huge fishing industry[16:50]
For some crustaceans extracted from the sea, much of the biomass is not consumable and becomes waste[17:08]
Current technologies for reusing these crustacean wastes often rely on very harsh chemicals[17:11]
They are exploring the development of enzyme-based technologies to treat these wastes[17:16]
Enzymatic processing could convert wastes into fertilizers or other useful products[17:25]
Using enzymes instead of chemicals would be more environmentally friendly because the reactions are carried out by biological means[17:30]
César mentions ongoing discussions with colleagues in Chile about pursuing this line of work[17:41]

Exciting developments in protein design

Miniature designed proteins as therapeutics

César highlights the work of fellow TED Fellow Chris Waltz and his company AI Proteins[17:59]
AI Proteins is developing very tiny proteins composed of few amino acids[18:49]
Very short proteins normally do not fold into stable shapes, similar to how a short rug cannot be folded[18:15]
AI Proteins has created tiny proteins that do fold upon themselves and have defined shapes[18:19]
These shapes are complementary to target cell surface proteins involved in disease processes[18:33]
Targeted diseases range from allergic reactions to cancer treatments[18:44]
The vision is to use these tiny proteins as pharmaceuticals (drugs)[18:49]
César suggests these protein-based drugs could be better solutions than various chemical compounds that lack such capabilities[18:55]
He says he is looking forward to seeing what AI Proteins does with this technology[19:04]

Future of protein design and enzyme engineering

Advances expected in the next 3-5 years

Lily asks where César sees protein design in 3-5 years and what might be possible[19:14]
César sees major advances in making new enzymes[19:37]
He explains enzymes are hard to design because they have specific sites on the protein surface where the reaction occurs[19:42]
He describes enzymes as like a sphere with a small hole on the surface that serves as the active site[19:51]
Chemical reactions occur only in this active site, requiring particular amino acids in very specific positions[19:59]
Methods available a year prior were described as "pretty bad" at enzyme design[20:09]
Previously, scientists would take a known enzyme with some desired activity and improve its sequence without changing its overall structure[20:10]
Newer AI-based methods can now afford changing the enzyme's structure and creating new structures from scratch[20:26]
This allows for the creation of new enzyme structures never seen before in nature[20:29]
Designing entirely new enzyme structures opens possibilities for new chemical reactions created from scratch[20:44]
César states the field is moving very fast toward developing new enzymes for new chemical reactions[21:43]

Risks, dual use, and regulation of AI in biodesign

Dual-use concerns and potential misuse

Lily asks about risks, dangers, and unintended consequences of designing proteins that nature has not made before[21:24]
César notes general concern about guardrails for AI technologies[21:08]
He explains that all AI technologies have dual-use potential: they can be used for benefit or for harmful impact[21:26]
Viruses are primarily composed of proteins and infect human cells[22:35]
With AI architectures for protein design, someone could potentially modify a virus to improve transmissibility or infection rate[21:41]
He classifies such malicious applications as harmful decisions[21:43]

Emerging safeguards and model evaluations

César says governments and companies are implementing approaches to assess the risks of these models[21:54]
Risk assessments involve different evaluations to understand potential misuse before releasing models to the public[23:02]
Scientists, including César and other prominent figures in AI for biology, signed responsible AI for biodesign guidelines[22:49]
These guidelines commit signatories to making significant efforts to identify risks in models developed for biodesign[22:57]
They also commit to indicating identified risks when releasing models[23:18]
Another suggested approach is "unlearning": adjusting models so they do not capture harmful potential when released[23:24]

Accessibility of tools and interaction with large language models

Currently, using these protein design models still requires expert scientists as they are not easy to use[23:33]
César warns that combining such models with large language models could enable non-experts to request harmful biological designs[23:37]
He gives an example of a person asking a language model to create a very harmful biological agent[23:54]

AI safety institutes and risk thresholds

César notes that the UK, US, and European Union have AI safety institutes[24:00]
These institutes evaluate the risk of using different AI technologies[24:06]
They use thresholds to determine whether a technology represents very high risk (requiring strong action) or very low risk (requiring monitoring but not strict oversight)[24:11]

Global landscape and leadership in AI-powered biodesign

Opportunities for countries beyond the US

Lily asks if shifts in global science policy create opportunities for other countries to step in and lead[24:57]
César believes there is opportunity for other countries to lead AI in biodesign[24:57]
He mentions Europe, noting that Denmark is heavily funding AI for biodesign[25:00]
The UK is also investing substantially in this area[25:05]
He notes efforts in Latin America to step into leading roles as well[25:35]
Chile has led an initiative for a while to assess risks and ethical usage of AI for various purposes[25:24]
He sees potential to use AI for protein design in Chile and become a leading Latin American country in that domain[25:48]
César suggests that as the US has undergone changes in the last year, many other countries have stepped up to lead in protein design[25:40]

Art, creativity, and communicating science

César's artistic background and personal history

Lily notes her own background in theater and film and asks César about his background in the arts[26:06]
César shares that his father passed away when he was seven[26:15]
Afterward, he struggled somewhat in school[26:30]
His mother enrolled him in many extracurricular classes, especially in the arts[27:03]
He began learning oil painting on canvas around age eight or nine[26:38]
He continued oil painting for several years, until around age 15[26:45]
He also learned to play the guitar during that period[26:52]
In high school, he became involved in acting[27:00]
For him, both arts and science are large spaces for creativity[27:16]
He sees both domains as places where one can push boundaries and expand horizons of what can be created[27:54]
He regards arts and science as similar playgrounds for exploring creativity's boundaries[27:52]

Using art to communicate complex science

César remarks that explaining his work to fellow scientists is relatively easy, but communication to the general public is complex[27:39]
He thinks arts can be a powerful tool to express very complex scientific topics[27:34]
He is considering what artistic renditions could help people understand what a protein is and what proteins do[27:44]
César believes that by connecting arts and science, it is possible to push boundaries even further than with either alone[27:52]

Hopes and fears about the future

Concerns after the COVID-19 pandemic

Lily asks what César is scared of and what is giving him hope[28:05]
César says that after the pandemic he often wonders what will come next[28:11]
He worries about the possibility of another pandemic and how people will respond[28:15]
He is more concerned about human behavior than about scientific capacity to respond[28:24]
He notes humans are very forgetful and may forget the experience of being in quarantine, as happened in Chile[28:26]
He does not want to be in that situation again[28:33]

Hope from biotechnological solutions to climate change

César is hopeful because of significant investment and scientific interest in biotechnological solutions to climate change[29:16]
He mentions plastic degradation as one example of such solutions[28:49]
He also points to efforts to eliminate greenhouse gases, using protein-based and cell-based solutions[28:59]
He notes there is strong interest in investing in these technologies[30:03]
César is optimistic that in the near future many biotech startups working on climate change will be successful[30:14]

Closing credits and TED Fellows information

Fellowship acknowledgment and where to learn more

Narrator identifies César Ramírez Sarmiento as a TED 2025 Fellow[29:36]
Listeners are directed to fellows.ted.com to learn more about the TED Fellows program and watch the films[29:41]

Lessons Learned

Actionable insights and wisdom you can apply to your business, career, and personal life.

1

Combining human creativity with powerful tools like artificial intelligence allows us to explore vast design spaces much faster than natural evolution, enabling targeted solutions to pressing problems such as pollution, disease, and climate change.

Reflection Questions:

  • • Where in your own work or life could you pair your creativity with a powerful tool or technology to accelerate progress instead of waiting for gradual improvement?
  • • How might reframing technology as a creative partner, rather than a replacement, change the kinds of problems you attempt to tackle?
  • • What is one concrete challenge you face today where you could deliberately bring in a computational or analytical tool to explore more possibilities than you could alone?
2

Building local capacity and communities around advanced technologies empowers regions to address their own specific challenges, rather than passively importing solutions designed elsewhere.

Reflection Questions:

  • • What challenges in your local environment or industry are not being adequately addressed by solutions designed in other contexts?
  • • How could you contribute to forming or strengthening a local community of practice around a critical skill or technology?
  • • What is one step you could take this month to share knowledge, mentor others, or collaborate locally so your community is better equipped to solve its own problems?
3

Any powerful technology, including AI for biodesign, is inherently dual-use, so responsible innovation requires proactively assessing risks, setting guardrails, and thinking through misuse scenarios before broad deployment.

Reflection Questions:

  • • What are the most significant ways your work or tools could be misused if they became widely accessible without context or safeguards?
  • • How might you incorporate simple risk-assessment and "red-team" thinking into your decision-making before launching a new product, feature, or project?
  • • Who could you involve-inside or outside your field-to help you see ethical blind spots or unintended consequences in your current plans?
4

Art and storytelling can make complex scientific or technical ideas comprehensible and emotionally resonant, dramatically expanding who can engage with and support important work.

Reflection Questions:

  • • Which parts of your work are hardest for non-experts to understand, and how might visual, musical, or narrative elements make them more accessible?
  • • How could you collaborate with artists, designers, or communicators to translate one complex concept you care about into a more intuitive form?
  • • What is one experiment you could run this quarter to explain your work in a novel, creative medium and get feedback from a non-technical audience?
5

Recent crises like the pandemic show that human behavior and collective memory can be as decisive as technical capacity, so preparing for future shocks means cultivating social resilience and learning, not just better tools.

Reflection Questions:

  • • In what ways did you or your community adapt (or fail to adapt) during recent large disruptions, and what lessons risk being forgotten?
  • • How might you capture and institutionalize the best responses you've seen-through protocols, stories, or training-so they're available when the next crisis hits?
  • • What small habit or routine could you establish now that would make you and those around you more resilient in the face of sudden change?

Episode Summary - Notes by Cameron

The new era of AI-powered protein design | César Ramírez-Sarmiento
0:00 0:00