From: redpointai
Eric Reese and Jeremy Howard are building Answer AI, a company focused on developing smaller, cheaper, and more affordable AI models and applications, particularly in the legal and education sectors [00:00:08]. Their approach emphasizes applying the Lean Startup process to AI [01:14].
Focus on Legal Applications
One of the broad areas Answer AI is excited about is law, given its heavily language-based nature [00:34:41]. As one lawyer described, a law firm functions like a very large language model: you feed in text and get out text [00:35:05].
The use of AI in law presents significant opportunities to help society [00:35:22]. Historically, the law has often been used as a “weapon” by wealthy individuals and organizations against less wealthy ones, creating grave injustices [00:35:31]. It’s also frequently used for gatekeeping in regulated markets, making it difficult for individuals to bring new ideas into the world [00:36:35].
By significantly reducing the cost of high-quality legal advice, AI can ensure that the law is less often monopolized by the powerful [00:36:12]. Answer AI has lawyers on staff working on the legal side of things [00:58:06].
Focus on Education Applications
Education is another key area of focus for Answer AI, also being a heavily language-based field [00:34:54]. Jeremy Howard, a co-founder of Answer AI, previously started fast.ai with the mission to make AI widely accessible, initially to non-PhDs, and to enable application creation with minimal code [00:48:52]. This mission continues with Answer AI, aiming to maximize the public benefit of AI for as many people as possible [00:09:39].
Jeremy highlights the many opportunities to improve education so that more people can realize their potential and build what they envision [00:37:22]. As a homeschooling father who has extensively researched academic literature and teacher interviews, he has observed the benefits of removing constraints from traditional teaching environments, where children often follow the same path at the same time with limited customization [00:37:06]. If AI can help more people access education through language-in, language-out systems, it would be transformative [00:38:08].
Underlying Philosophy and Approach
The R&D Lab Model
Answer AI operates as a for-profit R&D lab, a model that hasn’t been widely seen since the early days of electricity [00:14:54]. This model prioritizes coupling research directly with practical applications, a stark contrast to the hyperspecialization often seen in modern academia and industry, where research and development are often separated [00:15:35]. The belief is that the best research occurs when the researcher is closely connected to the application, allowing for continuous feedback loops from the customer back into scientific inquiry [00:16:02].
This approach is inspired by historical R&D labs like Thomas Edison’s, which Eric Gilliam describes as having a “Long Leash with narrow fences” [00:25:24]. This allows researchers considerable freedom within defined areas, fostering intuition about what the technology can and cannot achieve [00:31:38].
Importance of Cost Efficiency
A core tenet of Answer AI is to reduce the cost of AI models and their usage [00:26:18]. The current industry shows an “overinvestment” in training Foundation models from scratch and using expensive hardware, leading to an “underinvestment” in real-world, resource-constrained applications [00:25:40].
Reducing costs by factors like 10x is seen as a key to increasing accessibility [00:26:51]. This efficiency is not just about improving margins; it creates “a difference in kind,” making new applications possible that are currently unfeasible due to cost or supply chain constraints [00:27:17]. For example, if fine-tuning models becomes cheap enough, continuous fine-tuning of individual agents could enable “hyper-personalization” and memory in AI applications, which is not possible with current “amnesiac” models [00:28:45].
Connecting Research to Customer Needs
Answer AI emphasizes building practical, deployable products over splashy demos or speculative “science fiction” applications [00:29:54]. They believe that a significant reservoir of useful applications exists that do not require Artificial General Intelligence (AGI) [00:45:14]. These applications, often built with smaller, properly fine-tuned models, are intrinsically safer than relying solely on frontier models [00:46:15].
The company fosters a culture where team members are encouraged to deeply engage with the AI ecosystem and products, even by providing a monthly credit card for exploring new AI tools [00:33:11]. This hands-on experience, combined with deep understanding of the underlying technology, helps identify opportunities and constraints, informing both research and development efforts [00:34:10]. This focus on customer understanding and practical deployment is crucial for avoiding the “carnage” seen when companies merely “copy-paste” the SaaS stack to AI without understanding the unique economics and requirements [00:02:22].