The Last Technological Renaissance: A Conversation on AI with Peyman Faratin
It’s difficult to get through a conversation these days without the mention of AI. Yet while this transformative force is shifting thoughts, markets, and workflows, it's hardly a new concept… John McCarthy first coined the term back in 1956 when he called a meeting to discuss “how to make machines use language, form abstractions, and concepts.” Long before him, the likes of Alan Turing pondered on the subject. For decades, scientists and engineers ruminated on it, building on concepts, ideas, and expanding technologies that led to AI’s recent explosion and its seeming chokehold on modern society.
One of the key players in this arena is Peyman Faratin, a pioneer and intellectual leader who’s been working on artificial intelligence and machine learning for the last 35 years. Faratin is currently shaping the future of AI at Parameter Ventures, a venture capital firm purpose-built to help enterprise software pioneers create category-defining products. We sat down with Faratin to discuss AI investment, market predictions, and how companies like Perl Street can benefit from this transformative technology.
The Pathway to AI and Machine Learning
Peyman’s career objectives weren’t always so science-based. Upon immigrating to the UK in 1979, he originally aspired to be an artist. “I was going to be a painter,” he shares reflecting on his early career inclinations. However, Faratin's path took a significant turn when he decided that being “a single Iranian male in the Western world as a painter was not an option.” Jumping into cognitive psychology, Faratin built a neural network to discriminate between “br” and “pa” phonemes – his first exposure to AI. “The mind is viewed as a kind of a cognitive computational device. That was my first exposure to AI and computation.”
His academic and professional trajectory expanded from these early experiences into more complex fields. Faratin completed a master’s in cognitive science, contributing to projects that garnered significant attention. “For one of our projects, we implemented a neural a model of hippocampus, the region of the brain responsible for short and long term memories.. Researchers at Oxford had come up with a computational model of the hippocampus called the Hopfield network. When we implemented the model, we ended up showing empirically that the [Hopfield networks] theory didn’t hold. That got some attention.” Faratin explains.
After working on projects that touched on the subject, Faratin pursued a PhD in 1995 with a focus on AI. “I pursued a PhD in the field of distributed artificial intelligence and multi-agent systems, focusing on a paradigm where today’s computational systems are often designed and controlled by single, monolithic entities like Google, OpenAI, or Meta,” Faratin explains, “My thesis proposed that future AI systems would evolve away from this centralized model toward a more distributed architecture. Just as the internet transitioned from complex, centralized telephone networks to a system with a simple core and intelligent edge devices, AI will similarly shift to a more distributed approach.”
After his PhD, Faratin joined MIT’s Computer Science and AI Lab, teaming with prominent figures such as internet pioneer Dave Clark. Together, they worked to integrate AI into IP networks, “We were focusing on creating a 'knowledge plane'—an intelligent layer in the network architecture—beyond the usual data and control planes. The goal was to integrate intelligence into these networks.” Faratin says.
2007 saw the beginnings of Faratin’s entrepreneurial pursuits, starting with the co-founding of an early recommendation system and continuing on to a solo venture to tackle challenges in natural language processing (NLP). “We were working on NLP to address a problem: as social media emerged as a dominant source of content, it disrupted traditional NLP pipelines that were trained on clean, editorially curated data from sources like Wall Street or Reuters. Social media data, by contrast, is noisy and messy. Our approach in 2010 was to move beyond the atomic representation of words by expanding them into higher-dimensional vector spaces, a technique now known as embeddings.” That strategy transpired to be an important paradigm in moderl AI models In 2018, he joined Two Sigma, where he developed risk models and attempted to apply new AI methods, and has since joined forces with Parameter Ventures, a New York-based venture capital firm.
Role at Parameter Ventures
Parameter Ventures is dedicated to helping enterprise software pioneers develop category-defining products. For the past three and a half years, Faratin has been instrumental in developing the investment firm's strategy, leveraging his deep expertise in AI and technology. Faratin emphasizes the unique background of the team: “One of the key differentiators of the firm is we're very product and technology-focused and every partner has an operator background,” which provides them with insight into both product and technology. He continues, “We’re not just financiers, so we have a good understanding of how products are built.” This background enables them to add more than just capital, seeking to provide strategic guidance and support for their portfolio companies.
Faratin focuses on due diligence for Parameter Ventures, particularly in areas like AI, insurance, and Web3. He explains, “My primary responsibility is due diligence, especially technology due diligence.” This involves taking a deep dive into the technological principles of a product, understanding what makes it tick, and determining where it can fit in the market, which means assessing not only where a company is today but also where it will go tomorrow. Faratin notes, “A big part of my job is to assess how much of a company’s technology is just a thin layer on top of existing platforms and not truly defensible. There are many players in the AI space today who are merely adding API calls without substantial differentiation. It's crucial to identify where companies fall on the spectrum of competitive advantage and how they maintain it, especially given the sheer volume of inbound opportunities and the low cost of developing AI technologies today.” Faratin highlights the importance of a strong market strategy, a capable and experienced team, and a product that is defensible in the competitive landscape. "My role is to sniff that out," he notes.
Evaluating AI Investments
At Parameter Ventures, a few key elements are crucial when considering AI startups - and it’s not necessarily all in the technology. “Product, team, and tech are our main criteria, and the team is, of course, the most important at the early stages,” Faratin states, indicating that these are the essential factors they focus on. A solid go-to-market strategy is always important; Faratin notes, “It's not necessarily a deal killer if the revenue isn't there. What matters more is having a good team and a believable market strategy, along with a strong network. We're willing to overlook some revenue metrics if those other factors are strong."
Faratin stresses the importance of aligning a startup's vision with measurable outcomes. "The onus is on the entrepreneur to show that those activities are contributing to the bottom line," he notes. This pragmatic approach ensures that while innovation is valued, it must be grounded in a realistic business model that can sustain growth and profitability.
Perl Street: A Case Study
Looking at Perl Street’s offering, Faratin sees several opportunities for AI to make a significant impact. Perl Street stands out due to its focused approach to narrow AI applications, as opposed to broader, more generalized platforms, embodying a unique form of innovation that aligns well with the investment philosophy at Parameter Ventures.
In discussing Perl Street, Faratin highlights several key factors making the startup notable. He explains that AI applications span multiple layers, including workflow optimization and data harmonization. Perl Street is particularly relevant in this context, as it addresses the challenge of integrating separate systems and data sets within organizations. Faratin elaborates, "Organizations have historically evolved into collections of separate systems and data sets, with different business units operating independently." He contrasts this with the traditional approach, which aims to create a unified global data model. With foundational models, "we have tools to address this issue from the bottom up," allowing for a shift from a top-down approach to one that handles data complexity through flexible computation rather than brittle designs.
Faratin further explains, "Much of the work in AI and machine learning involves this 'janitorial' data work, which is essential but often overlooked." He underscores that using advanced AI tools for data harmonization is crucial for effective AI deployment, as it lowers transaction costs and enhances efficiency. Perl Street's focus on optimizing workflows and harmonizing data exemplifies this critical step in AI development.
Reflecting on his experience at Two Sigma a New York City-based hedge fund, Faratin mentions that AI can significantly improve risk modeling, such as in portfolio optimization and project finance, by enhancing predictions with harmonized data. “Another value is the ability to model risk. Portfolio optimization and project finance are fundamentally about understanding and mitigating uncertainties. AI plays a crucial role here, as demonstrated by my work at Two Sigma, where we used it to model risk,” he shares.
Future Global Trends in AI
Faratin highlights several key trends shaping the future of AI. One significant trend is the improvement in data efficiency, particularly in areas like computer vision. He explains, "Today’s models are technically sample inefficient. They require so many data points in order to learn basic concepts." Faratin elaborates on the advancements in data augmentation, stating, "What that community has done very well for the past number of decades is you take a data set that basically was for data augmentation... expanding the data set."
Another trend is the influence of robotics. Faratin notes, "The rise of robotics is going to be really interesting," and he emphasizes the role of foundational models in accelerating this field, saying, "Transformers are actually very applicable... [and] have become a de facto standard in our field." This common platform allows researchers from different AI domains to communicate more effectively: "A language researcher can now talk easily with a vision researcher, with a speech researcher," he observes.
Faratin also discusses the future integration of robotics and AI, describing a "virtuous cycle" where physical interactions with the world enhance AI's perceptual capabilities. He illustrates this with an example: "When your child holds an object, they don't see the pixels on the backside of it, but their hand provides sensory information about what they can't see. This concept, rooted in developmental psychology, illustrates that behavior is an integral part of perception. Today, we have advanced perceptual machines, but once these machines interact in real-world environments, they’re going to provide a huge amount of information in a very efficient manner."
Faratin's emphasis on handling ambiguity and making informed judgments is essential in an area as dynamic as artificial intelligence (AI), where revolutionary developments are continuous. "It's about controlling risks while considering what we know, managing uncertainties that come with the combination of innovation and practical market realities," he says in summarizing his strategy. Faratin's viewpoint, which is based on decades of experience, shows a thorough comprehension of the development of artificial intelligence, from its early conception by pioneers like Alan Turing and John McCarthy to its current function in influencing contemporary culture. His goal is to strike a balance between cutting-edge research and useful application, making sure that artificial intelligence (AI) keeps advancing diverse industries while tackling both the fascinating opportunities and the practical issues it poses.