The primitives of AI
David Peterson

Back in 2009, Albert Wenger from Union Square Ventures wrote this piece on the opportunity in mobile. In it, he argued that the most exciting opportunities will be those that are “native” to the platform and take advantage of mobile-specific primitives. Those primitives, according to Wenger, were: location, proximity, touch, audio input, and video input.

In an effort to keep looking backward to move forward, I’ve been thinking about what the primitives for AI technology may be, to see if that might help me think through what opportunities are most exciting in AI.

Here’s a starter list that I’ve come up with:

• Generation — text, image, audio, video. Each week, it seems, we see increasingly impressive advances in content generation across the board. Sora (OpenAI’s video generation model) was just the latest. One exciting (if a bit uncomfortable) fact about AI models is that they are eerily good at mimicking humans (so much so that I debated adding “mimicry” as a separate primitive). Anybody can experience this by prompting one of the foundational models to generate content “in the style of” a particular person. But there are countless tools out there based on leveraging mimicry alone to create novel experiences…tools that let you chat with a specific person, test a product with somebody who looks like your average user, survey someone who looks like a likely voter, and so on.

• Hallucination — hallucinations are a completely novel type of response from an AI model, and as Andy Weissman wrote here, they may indeed be a “feature, not a bug” of AI. I think of hallucination as the native ability to add randomness. The better we get at controlling hallucinations, the more we can leverage AIs hallucination capability as a “randomness dial” that we can turn up and down depending on the use case. Randomness may be useless (and even harmful) in some scenarios, but evocative and catalyzing in others.

• Retrieval & Memory — this was already getting to be true with advances in RAG, but as was made clear in the Gemini 1.5 paper released last week, AI is going to get really, really good at near-perfect information retrieval, even from a massive, multi-modal corpus, very soon. While the underlying technology will likely be different, near-perfect information retrieval looks an awful lot like near-perfect memory recall, so I’ve put these two primitives together in the same category.

• Multi-modal input — and last but not least, the latest AI models can now understand text, image, audio and video input, making them nearly as flexible in understanding as a human counterpart.

(Note: one primitive I did not mention, but I imagine is coming soon, is action. Take a look at the “large action model” from Rabbit as an example for how a startup, at least, is approaching this problem.)

What makes this moment particularly thrilling is that these primitives, unlike during the mobile era, are themselves brand new. The mobile era was exciting because there was a convergence of existing capabilities in one device. The AI era is exciting because a whole set of new-to-this-world capabilities are converging with the capabilities we already have. I don’t know which AI native applications will emerge as winners, but I’m particularly interested in those companies that do the best job of marrying the new and the old.

As an example, we already have tools that are pretty good at answering analytical questions (e.g. coding languages). With AIs native generation and retrieval capabilities, it would seem that someone could build a 10x better research and analysis platform. If you think about how you complete research tasks today, you likely spread your work across myriad tools. Lots of searching, reading, note-taking, cross-referencing, reading again, and so on. It’s iterative, cyclical and creative. I had a professor who said good research is hard because it requires unbounded exploration while keeping your feet planted firmly on the ground. This seems like an ideal task for a model with near-perfect memory and information retrieval over one set of content and inhuman synthesis capabilities across every piece of content ever produced. The particular feature set will depend on the target customer, but this sort of platform would seem particularly valuable for industries where the upside to more/better/faster ideas is clear. Perhaps a hedge fund analyst searching for alpha?

As I’m imagining what that product might look like, I can’t help but think that by mashing together new capabilities with the old, you end up with a product that feels plucked from science fiction and almost unreal. It’s too much like a companion, not enough like a piece of software. But I think that’s the world we’re entering.

I’ll keep riffing on these ideas over the next few weeks. But would love to hear your feedback. Which primitives did I miss? What application categories are most interesting to you? And if you’re building something that’s bringing together the old and the the new…I’d love to hear about it.

David

FROM THE BLOG

A Digital Fabric for Maritime Trade
Why we invested in Portchain.

Three Keys to the Kingdom
The sometimes-competing and sometimes-aligned goals that early-stage founders must manage.

Customer-driven Entrepreneurship
Reframing the critical unlock in early-stage venture.

It’s 3am. Do You Know What Your Third-party API Calls are Doing?
Announcing our investment in Lunar.dev.

EUROPE AND ISRAEL FUNDING NEWS

UK / RegTech. Napier raised £45M, led by Crestline, for its AI-enhanced financial regulation compliance and AML software platform.

France / Generative AI. Bioptimus raised $35M, led by Soffinova Partners, for its foundational model for bio.

France / SaaS. AZMed raised $16.2M, led by Maison Worms, for its AI-powered platform to assist radiologists identify fractures (and soon other medical issues as well).

Israel / DevTools. Faddom raised $12M, led by Viola Ventures, for its IT infrastructure and dependency mapping platform.

Denmark / SaaS. Go Autonomous raised €10M, led by Octopus Ventures and Ridge Ventures, for its B2B commerce automation platform.

Denmark / SaaS. Spektr raised $5.4M, led by Northzone, for its compliance management and due diligence orchestration platform.

WORTH READING

ENTERPRISE/TECH NEWS

Palo Alto Network’s tough quarter. The security vendor’s shares were sharply down this week following a weak quarter and, more significantly, the announcement of a new strategy — seemingly designed to counter a weakening of demand for cyber products from overwhelmed CISOs. “Palo Alto’s new strategy includes an aggressive marketing approach based on a sort of “trade-in” transaction. This will allow organizations to “trade-in” old security systems from various suppliers in the cyber market that are still under contract and to switch to Palo Alto’s product packages free of charge for the first six months. The presentation of the new strategy sharpened the feeling that Palo Alto was experiencing its Microsoft moment, and not in a positive context. Just as Microsoft steamrolled many competitors when it pushed free products through its operating system, Palo Alto now seems like it intends to do the same. As evidence, at the beginning of today’s trading, the likes of SentinelOne, Check Point, Crowdstrike and Zscaler all suffered significant drops in their share price. In the conference rooms of the cyber startups, most of them Israeli, they must have been sweating yesterday in the face of Palo Alto’s new statements about “platform consolidation” and “customer fatigue.””

Intel in Israel. Calcalist reports on how Intel CEO Pat Gelsinger’s plan to separate development and production will impact the company’s activities in Israel. “Intel has been manufacturing chips on a limited scale for companies like ARM and even Nvidia for several years but now Gelsinger is separating manufacturing and development to prevent structural conflict of interests with manufacturing becoming an independent profit and loss unit. Intel Foundry will include IFS, the fastest growing production unit in Intel in 2023 with revenue of nearly $1 billion. The division will include all the group’s existing manufacturing fabs worldwide such as the chip plants in Kiryat Gat and its flagship plant in Oregon, which is responsible for proving the programming of new production technologies.”

Will AI remove humans from supply chain management? This article on Hackernoon argues the case. “All but 6% of supply chain businesses already use AI in some capacity. Roughly 11% say it’s critical to their operations, a figure that could apply to more than a third of these companies before long. It’s easy to understand why as AI improves many aspects of supply chain management. AI’s predictive capabilities are one of its strongest assets in supply chain management. Machine learning models can analyze past sales data and current trends to predict upcoming demand shifts. Companies can then order less of some products and more of another to prevent stock-outs and surpluses.”

HOW TO STARTUP

AI is like water. Morgen Beller at NFX wrote a thoughtful piece on how AI is like water: necessary, ubiquitous, and always the same. His conclusion is the tech differentiation is rare and getter rarer: “For a long time, tech has been a differentiator among software startups. We thought that if you could build something no one else could, that would be enough to protect you — definitely not forever, but at least for a while. But what we’re seeing with AI is that tech provides you basically no protection from the start. Tech differentiation in AI is a shrinking moat.” Beller argues that just as companies like Perrier used marketing to build the perception of value, tech companies will need to lean more heavily on marketing strategies as differentiation goes to zero. ”It’s not a nice takeaway, but it’s the reality. You can either guarantee a loss and not play the brand and marketing game. Or, you can give yourself a small chance of success by playing the game, and seeking other avenues to build your advantage.”

Are SaaS Decacorns back? Jason Lemkin highlights that a number of SaaS giants have crossed back into decacorn territory. These include Monday, Confluent, Procore, and Gitlab. “Bottom line, there are more SaaS and Cloud leaders at the dark blue line of a $10B+ market cap than there has been in quite a while. It doesn’t mean it’s easy. To be a decacorn in 2021, it took a lot of growth, but not really any efficiency or profits. Today, it generally takes $1B+ ARR, strong growth AND a very efficient engine. The trifecta. Much harder.”

HOW TO VENTURE

The best asset class there is. Packy McCormick offers an impassioned defense of the societal virtues of the venture asset class. “The invisible hand is more visible in venture capital than it is in any other asset class. Venture is an ecosystem made up of parties acting in their own self-interest that seems to operate with some sort of collective intelligence on a longer timescale… The opportunity to generate returns today might not exist if someone hadn’t been willing to lose money decades ago. No venture capitalist does this altruistically — they’re driven by the small, against-all-odds chance that this too-early technology might be the next big thing — but in their mistakes they create opportunity for others.”

Are US VCs struggling in Europe? Some people think so, as Techcrunch has reported. From our vantage point, this data is far from conclusive, but here is what Techcrunch is arguing: “Big names like Coatue and OMERs formally pulled out of the region in recent months, and the venture funds that have remained are significantly less active. Navina Rajan, a senior analyst at PitchBook, said that the overall value of European deals with at least one U.S. investor declined 57% in 2023 compared to a year earlier, and deal count declined 39%. To compare, overall deal value declined 46%, and deal count declined 31% in the same time frame. The European startup market comes with nuances that make it a difficult one for North American investors. Each country in Europe comes with its own language and sometimes currency. Investing in both Romania and Italy is different from investing in both Texas and California. Plus, startups and universities produce different networks for European startups than in the U.S. Taken together, all of those nuances make for a challenging market in the best of times, let alone the doldrums of the past couple of years. It’s no wonder then that North American investors have struggled to find a secure footing as they try to straddle the Atlantic.”

What went wrong at Techstars? The decision by Techstars to shut down its Seattle operation seems to have sparked a backlash. In this piece, Chris DeVore argues that the renowned accelerator went astray by chasing corporate sponsorship dollars to keep itself afloat. “Bottom line, Techstars needed cash. And since the program-based fund model didn’t provide it, Techstars started looking within the local ecosystems in which it operated. It began with an effort to extract “sponsorship” dollars from local service providers: the lawyers, accountants, recruiters and PR firms that cater to startups. But for every sponsor who agreed to write a check, there were a dozen other vendors who were equally or more qualified to support the teams in the program. What did we owe our sponsors, and did that put us in conflict with our commitments to give founders the best possible advice, and to never waste their time? The next logical step was to go up-market and look for financial “partners” among the many corporations struggling to keep up with the pace of technological innovation during the go-go ZIRP years. Techstars was attracting many of the world’s best founders, surely some of those founders would be interested in solving problems faced by these large corporations? It’s not hard to see where this all leads: from a principled beginning, laser-focused on helping the world’s best founding teams achieve the best possible business and financial results, soon the Techstars system began to play host to mandatory, sponsor-led “education” sessions for participating teams. Next, entire accelerator programs were created on behalf of corporate partners, promising them access to cohorts of world-class founders eager to listen to their needs and use their APIs.”

PORTFOLIO NEWS

Januar raised €1.5M seed extension and launched Januar Crypto — an institutional wallet and crypto trading solution integrated alongside Januar’s fiat account and payment solutions.