David Friedberg on AI-First Startups & the Future of Biology, Business & Creativity | AI Basics

Key Concepts

AI leverage, plant breeding, genome language models (GLM), automation, venture capital, deep tech, rare earth minerals, data normalization, hallucination in AI, AI-driven media, cultural storytelling, abundance, digital biology, DNA sequencing, phenotype prediction, AI-first culture.

Main Topics and Key Points

AI's Impact on Startup Operations

• Leverage and Efficiency: AI is revolutionizing how startups operate, enabling smaller teams to achieve more. Every job, function, and workflow can be improved, leveraged, and scaled using AI.
• Resume Screening Example: David Friedberg's chief of staff, Laura, used AI to write a Python script to download, scan, and score hundreds of resumes from Greenhouse in a few hours, a task that would have taken weeks to outsource or many hours to do manually.
• Project Planning Example: AI was used to create a project plan for controlled environment experiments with LED lighting by feeding research papers and guidance to the AI, completing the task in hours instead of weeks or months.
• M&A Transaction Analysis: Jason Calacanis used Google's Notebook LM to analyze M&A documents, cap tables, and correspondence, enabling a junior analyst to ask informed questions of opposing counsel, saving significant time and legal fees.
• Team Size and Deep Tech: AI allows smaller, nimbler teams to tackle complex, deep tech projects that historically required large companies with hundreds or thousands of employees.
• Venture Capital: The addressable economic opportunity for startups has grown because AI provides leverage, allowing smaller teams to pursue multi-billion dollar markets.
• Rare Earth Mineral Extraction: AI can be used to design automated drilling machines and project plans for extracting rare earth minerals, addressing complex technical challenges. The energy cost is a square of the depth you go when you do mining.

Cultivating an AI-First Culture

• Bifurcation in Companies: There's a divide between employees who are AI-first and those who are not.
• Lunchtime Seminars and Hackathons: Oho Genetics conducted lunchtime seminars and hackathons to teach employees how to use AI tools like Cursor and apply them to business problems.
• Challenging Headcount Requests: Managers should challenge employees to justify why AI cannot be used instead of hiring additional staff.
• AI-First Offsites: Leadership teams should participate in offsites to systematize how the organization will work with AI and define expectations for employees.
• Leading by Example: Leaders must be AI-first and disseminate processes throughout the organization with checks, accountability, and performance standards.
• Framework for Internal Use: The framework used internally is "automate, AI, deprecate, delegate".
• Job Description Requirement: Require every job description to include a paragraph explaining why AI cannot perform the job.

AI's Impact on Biology and Plant Breeding

• Digitization of Biology: Biology is being digitized, turning it into data that AI can analyze.
• DNA Sequencing Cost Reduction: The cost of DNA sequencing has decreased dramatically, from millions of dollars to under $100 per human genome.
• DNA Sequencing Process: DNA is amplified, chopped into short sequences, read by a DNA sequencing chip, and then statistically reconstructed to assemble the complete genome.
• Phenotype Prediction: AI can predict complicated phenotypes (physical characteristics) of plants by analyzing combinations of genes.
• Plant Breeding: AI can identify the perfect DNA sequence for desired plant traits (e.g., tall, drought-resistant, high-yield).
• Microbiology: AI is used to program E.coli cells to produce specific molecules, such as biologic drugs that target cancer cells.
• Genome Language Models (GLM): DNA sequences are treated like a language model, where AI predicts or scores a DNA sequence based on a specific objective.
• EVO2: The Arc Institute's open-source model, EVO2, acts as a grammar checker for DNA, predicting the functionality of a gene.
• Ohalo 2 Site: An archaeological dig on the Sea of Galilee revealed evidence of farming and seed selection dating back 26,000 years.
• Selective Breeding: Historically, plant breeding involved selecting the biggest plants and planting their seeds, leading to gradual improvements in yield.
• In Vitro Breeding: Accelerating breeding cycles by conducting the entire process in a petri dish.
• Dwarf Wheat Project: Norman Borlaug's dwarf wheat project, which took 20 years, could now be accomplished in a year or two using AI.
• Custom Strawberries: The idea of creating custom strawberries with enhanced nutrients, such as vitamin D, was discussed.
• Seed Production: Oho Genetics is working on enabling seed production in crops that are traditionally farmed vegetatively, such as potatoes.

The Era of Abundance

• Abundant Resources: We are entering an era of abundant energy, food, water, and resources.
• Water Desalination: If energy production costs drop to near zero, salt water can be converted into regular water.
• Element Creation: Theoretically, any element can be created if energy costs are low enough by using heavy ion beams to bombard atoms.
• Automation: Robots can automate housing construction, road building, and other tasks, with the cost primarily determined by electricity.
• Calorie Production: We currently produce 3,500 calories per person per day on Earth.

Addressing Hallucinations in AI

• Models of Models: Using multiple models, running outputs in parallel, and incorporating chain of thought with QA/QC can improve accuracy.
• QA/QC Agents: Employing AI agents to verify and confirm information.
• Multimodel Output: Using multimodel output and only accepting results where all models agree.
• Application Layer Solutions: Application layer and foundation model companies are building in checks and multimodel layers to resolve better outputs.

AI and the Future of Media

• Dynamic Media Creation: AI can dynamically create media, such as movies, based on user preferences.
• Artist's Role: Artists can define what is variable and what is not in AI-generated content, setting constraints and boundaries.
• Shared Cultural Experience: The core elements defined by the artist become the shared cultural experience, while individual rendering and interpretation vary.
• Democratization of Creativity: AI can enable more people to experience creativity and participate in making movies and other art forms.

Notable Quotes or Significant Statements

• David Friedberg: "Every job, every function, every workflow can be improved, levered, scaled up using AI."
• David Friedberg: "We can address tougher problems than we historically think about addressing."
• David Friedberg: "The GLM is effectively predicting or scoring a DNA sequence based on some objective."
• David Friedberg: "I do think we're entering an era of abundance abundant energy abundant food abundant resources abundant everything water."
• Jason Calacanis: "What does it mean to be AI first?"
• Jason Calacanis: "Have you tried to automate this? AI? Have you deprecated?"

Technical Terms and Concepts

• CRISPR: A gene-editing technology used to modify DNA sequences.
• Phenotype: The observable characteristics or traits of an organism.
• Genome: The complete set of genetic material in an organism.
• DNA Sequencing: Determining the order of nucleotide bases in a DNA molecule.
• Base Pair: A pair of complementary nucleotides in DNA (A-T, C-G).
• Gene: A unit of heredity that is transferred from a parent to offspring and determines some characteristic of the offspring.
• GLM (Genome Language Model): An AI model that predicts or scores a DNA sequence based on a specific objective.
• LLM (Large Language Model): A type of AI model that uses deep learning algorithms to process and generate human language.
• In Vitro Breeding: Accelerating breeding cycles by conducting the entire process in a petri dish.
• Hallucination (in AI): The tendency of AI models to generate incorrect or nonsensical information.
• QA/QC: Quality Assurance/Quality Control.

Logical Connections

The discussion flows from the practical applications of AI in startups to its transformative potential in biology and agriculture. The conversation then shifts to the broader implications of AI for resource abundance and the future of media, culminating in a discussion of how to address the challenges of AI hallucinations.

Data, Research Findings, or Statistics

• DNA sequencing cost has decreased from $100 million to under $100 per human genome.
• Oho Genetics is working on a system to replace 2.5 tons of leftover potatoes per acre with less than 10 grams of seed.
• The Earth currently produces 3,500 calories per person per day.

Synthesis/Conclusion

AI is rapidly transforming startups, enabling smaller teams to tackle complex problems and achieve greater efficiency. The impact of AI extends beyond software and automation to deep tech, biology, and agriculture, with the potential to revolutionize plant breeding, drug discovery, and resource management. As AI continues to evolve, it will be crucial for companies to cultivate an AI-first culture, address the challenges of AI hallucinations, and harness the power of AI to unlock an era of abundance and creativity.