«Большинство бизнесов провалит ИИ-трансформацию». Андрей Дороничев — о будущем технологий

AI-Driven Drug Discovery at BIOPTIC

Andrey Doronichev describes BIOPTIC, a startup that uses AI to identify candidate drug molecules. Its models encode possible molecules and use scientific data about molecule–protein interactions to predict whether a molecule may affect a specified protein target.

The company says its system is target-agnostic: it can search for candidates against targets it was not specifically trained on, rather than being limited to one protein or disease.

A collaboration with Stanford researchers provided early validation. Stanford supplied a specialized testing setup using cells derived from patients with a rare disease. Of five AI-generated molecules tested, several showed activity, and two reportedly performed better than the strongest previously known candidate for that target. The work led to a joint patent—BIOPTIC holds 60% and Stanford 40%—and a paper in Cell. These are promising early-stage results, not a proven treatment or cure.

Doronichev emphasizes the limits of current AI drug discovery. Finding an active “hit” is only the first step: candidates must be optimized, tested in increasingly complex biological systems, and eventually evaluated in clinical trials. Each stage can take years and cost millions or more; clinical development can reach very high costs. AI can search vast virtual chemical spaces far faster than traditional lab screening, but it cannot yet simulate a whole patient reliably or eliminate the need for real-world testing.

BIOPTIC’s Business Model

The company has two related paths:

• AI software for pharmaceutical work: BIOPTIC’s agent is presented as a specialized AI assistant for pharma, likened to a coding assistant for software developers. It can help automate research and business-science workflows. The company charges for usage, including tokens processed.
• Drug-development intellectual property: BIOPTIC can patent promising molecules and potentially earn royalties, license the intellectual property, or sell rights to a larger pharmaceutical company. Doronichev describes this as a long-term, high-risk route; in biotech, an IPO may fund further development rather than represent a final exit.

Doronichev says that large pharmaceutical companies have begun working with the startup, despite its smaller scale. In his view, established firms may find it difficult to rebuild their processes around AI, while smaller technology companies can move faster.

AI Adoption and Business Transformation

Doronichev predicts that many large businesses will fail to transform around AI. Their size, existing processes, and employees’ incentives can create resistance—especially when AI automates work performed by people inside the organization. Smaller, AI-first companies may be better positioned to build new workflows without that organizational baggage.

He describes the AI industry as a stack: energy supply, chips, data centers, foundation models, and then applications that address specific business problems. He sees the hardest challenge as the “last mile”: getting AI tools adopted in real organizations and embedding them in day-to-day work.

In his account, the biggest infrastructure investments may outlast a market correction, even if some current valuations prove unsustainable.

Lessons from Google Products and Innovation

Doronichev recounts several technology projects and what they illustrate:

• YouTube on mobile: He joined Google with the aim of making YouTube a mobile app, anticipating that video consumption would shift to phones. He says mobile later became the majority of YouTube viewing.
• Google Cardboard and VR: A small team developed an inexpensive cardboard viewer that used a smartphone and simple lenses to create a basic VR experience. The project grew from a prototype into a widely distributed entry point to VR. Doronichev credits Google’s protected “20% time,” tolerance for experimentation, and support for small projects.
• Stadia: Google’s cloud-gaming effort demonstrated that strong technology alone is not enough. Doronichev says Stadia struggled because it lacked compelling games and entered markets where many customers already owned established gaming platforms.
• A YouTube mobile redesign: A redesign that let viewers minimize the video player while browsing initially reduced watch time substantially. The team diagnosed the problem, tested many fixes, and found that a change in which browsing panel appeared first had hidden recommended videos. After adjustments, viewing time recovered and surpassed its earlier level. The example illustrates the risk of major product changes and the need for rapid iteration after launch.

He distinguishes continuous innovation—incremental improvement—from disruptive innovation, which requires a major change in direction and may initially perform worse. His broader lesson is that breakthroughs can be pursued systematically: build capable teams, create protected time to explore, tolerate failure, and be ready to act when an opportunity appears.

Guides, Reviews, and Practical Frameworks

There is no step-by-step technical tutorial in the video. The discussion does offer several practical frameworks:

• Treat major life or business changes like product development: define what matters, form a vision, and test it through small, reversible steps before making an irreversible commitment.
• Distinguish reversible decisions, which can be made quickly and corrected, from one-way decisions, which require more preparation and evidence.
• Expect a disruptive project to pass through an initial excitement peak and a period of disappointment. After launch, sustained testing and incremental fixes are often more important than the initial leap.
• To learn a new field, Doronichev describes building a network of practitioners, taking structured lessons, reading research, and using AI tools to help organize questions and material.

Main Speakers and Sources

• Alexander Sokolovsky — Host and interviewer.
• Andrey Doronichev — Guest; entrepreneur, former Google executive, and BIOPTIC co-founder.
• Stanford researchers and laboratory — BIOPTIC’s research collaborators and experimental validation partner.