By Innospark Ventures
2023 was a transformative year for AI. Far from being incremental, the year was marked by an almost instantaneous gain of the mainstream mindshare by AI. Our Founder Venkat Srinivasan reflected, “LLMs had the black swan moment of 2023, and we couldn’t have predicted how quickly they consumed the common man and nation states alike.” Major tech companies invested heavily in AI startups, policymakers around the world began to grapple with AI regulation, venture capital firms rushed to claim AI expertise, and startups leaned heavily on LLMs in exchange for traditional AI/ML approaches. But through all the hype, 2023 lent itself to significant innovation and opportunity.
After a fast-paced start to 2024, the Innospark Ventures team took a moment to look back on this past year, and to consider the year ahead. We’re sharing our insights on the most surprising trends, our thoughts on the major technological developments we expect to see this year, and predicting what ethical considerations and regulatory frameworks will develop in 2024.
Trends and Surprises. The most unexpected trends and developments of 2023, and how they influence our investment strategy in 2024
I was surprised by the speed at which the healthcare systems ‘moved’ in developing strategies to adopt generative language models. Though the prospect of using LLMs in treating human health is undoubtedly exciting, it was still surprising to see large (& slow) bureaucracies move so quickly on new technology. Further, I was even more surprised to see systems forge partnerships with startups in addition to the legacy providers, who so clearly have the advantages of deep relationships, massive installed bases, integrated systems, regulatory compliance, and known UI/UXs. This, to me, says there is plenty of innovation ‘left’ for startups to compete on product and technology with big tech and EMR companies. It also speaks volumes about the problems systems and providers face with which AI could be uniquely situated to help combat: burnout, staffing shortages, increase in patient communication, increase in paperwork, data overload, overrun ERs—to name just a few(!). – Mark Legare
I was pleasantly surprised by the breadth of AI applications reflected in life science industry dealmaking. For an industry not always considered an early tech adopter, we saw deals reflecting AI’s incorporation into several areas of the discovery and development process - from target discovery (ex. BI + Phenomic, AZ + Verge), to hit identification (ex. Merck + Benevolent, Lilly + XtalPi), to lead optimization (ex. AZ + AbSci), and even to clinical trial documentation (ex. Novartis + Yseop). I’m entering 2024 encouraged that the industry will continue engaging this transformative technology in its mission of getting effective treatments to patients faster. – Jeff Knox
I was surprised by the focus on generative AI in 2023. Sure, after the launch of ChatGPT to the public, I too expected LLMs to be at the center of AI talk for the foreseeable future. Everyone, from my 16-year-old cousin to my 70-year-old aunt, suddenly started talking about it and using it in their daily lives. But when thinking back a few years to when I first joined Innospark, my answer to the question 'What do you think AI will be all about in 2023?" would not have been "GenAI and LLMs". I might have said something along the lines of autonomous vehicles or quantum computing. So, in the grand scheme of things, the major AI themes of 2023 were a big surprise. – Max Krause
Tech Predictions: The most significant developments we can expect to see in 2024
The current LLM architecture will be replaced by an entirely new architecture which will be explainable, transparent and require a lot less compute. - Venkat Srinivasan
As we approach 2024, I believe the true AI experts (those who have been actively involved in the field before the introduction of ChatGPT, academics with commercial interest, exceptional technologists, etc.) and founders with deep domain expertise will come out on top. Regrettably, lacking either of these qualities will make it exceedingly challenging to compete with tech giants, who wield a vast array of capital and resources, and face stiff competition from numerous other founders entering the AI space. – Eon Mattis
I anticipate we will see a proliferation of products that claim to address the limitations of LLMs (hallucinations, tractability, data privacy etc.), and in some cases I expect they will have made real progress within specific use cases or verticals. I am hopeful that founders will focus on solving real problems, that they understand deeply, and leveraging potentially disruptive tools like AI models to reframe their solutions vs. looking for problems to solve with LLMs. – Matt Fates
AI Ethics, Regulations, and Oversight: How will these evolve and impact the pace and direction of AI innovation in 2024
Regulators will publish clearer rules and acceptance criteria for AI based models; stipulate how AI will be developed and used; this will have a major impact on who develops AI and its evolution. This will remove uncertainty for startups applying AI to solve problems and be a net positive for investors. - Venkat Srinivasan
We’re living in an unquestionably volatile time be it world events, turbulent markets, domestic politics, or climate concerns. Unsurprisingly, we’re seeing AI’s implementation sit neatly among the fault lines of how different groups view personal privacy and freedoms. There’s not a starker contrast than the EU’s new AI Act versus China’s approach of choosing who wins, based on party co-optation and managed outputs. I hope that we will see a counterapproach emerge to the ‘scrub the internet’ data arms race where more models are built using smaller, more targeted data sets. Thus, balancing good technology with basic human privacy rights. Going forward, I expect the US government to signal its stance on “what is possible” versus “what is right” in how it elects to adopt new technology, e.g., using AI for standard data-processing work (e.g. IRS forms) versus more controversial areas (e.g., policing). – Mark Legare
As a major election year approaches, I hold a contrarian view that there might be limited regulatory developments in the US concerning AI. The nation's citizens and political figures are likely to be intensely concentrated on selecting a president amid the turbulent times we currently face. While presidential candidates may articulate plans to regulate the growth of AI, the demands of campaigning are likely to consume their attention. - Eon Mattis
In 2023, US policymakers increased their focus on AI. Congress introduced the CREATE Act to enhance AI access for students and researchers, aiming to expand development opportunities. In 2024, I expect to see action on initiatives like the CREATE Act, but progress on comprehensive regulations to harness AI technology will lag. As legislators become more comfortable with the topic, and stakeholders increase lobbying efforts (AI lobbying up 185% in 2023) the debate will become increasingly partisan. Without unified federal guidelines, states will continue to establish their own AI regulations for discrete policy concerns like consumer privacy, creating a fragmented landscape that challenges startups and established entities alike. This will shift innovation to regions with more favorable regulatory climates, influencing which states will lead in AI innovation and talent going forward. – Lily Zarrella