GenAI Summit: An Ushur Intern's POV
July 30, 2024
Get the best, coolest, and latest posts delivered to your inbox.
Between May 29th-May 31st, I attended a major tech conference in the Palace of Fine Arts in San Francisco, California. The GenAI Summit SF ‘24 highlighted the modern era of generative artificial intelligence, or Gen AI. During the event, AI-first companies were given a chance to showcase both their products alongside their entrepreneurial journeys, and I was given the chance to attend and learn.
As a marketing intern, I felt fortunate to attend the summit. It was a very immersive experience. Almost everywhere I turned, attendees were excitedly talking about GenAI applications. In fairness, what better location than the Bay Area to host a tech conference? You could certainly feel the energy and excitement.
I spent the first day getting my bearings and planning which sessions to attend throughout the week, though I also got to speak with visitors at the Ushur booth. That meant I either was talking to people about Ushur's latest launch or attending seminars with knowledgeable experts in the field of artificial intelligence.
It felt like I was thrown in the deep end, but I was excited for the chance to talk to AI builders, developers, professors, and executives.
While the summit program contained a series of speakers, one of the most notable to me was a company founder named Jake Heller. Jake explained the ideas behind his company, CaseText, which was recently acquired for a $650 million dollar deal with Thomas Reuters.
CaseTexts’s product “Co-Counsel” is an AI legal assistant. It was trained in the skills of a legal assistant so that it could accomplish more menial and time-consuming tasks like summarizing legal documents, generating research based on prompt questions, and even reviewing evidence of fraud.
Heller talked about the impact his product has already had acting with the same skill sets as a legal assistant as opposed to just another computer. In his eyes, typical practices regarding legal affairs were traditional and outdated. Technology wasn’t nearly being utilized as much in legal practice as opposed to other industries such as insurance or healthcare. As a result, he felt the need to find an innovative and modern solution that would cut down time for attorneys to handle the most important tasks.
In the CaseText approach, Generative AI models were trained with a set of specific capabilities in mind but still used flexible large language models. That approach made sense and seems clever.
Similar to Heller, several other entrepreneurs at the conference told their own stories of how they grew their business, all of which were by different approaches.
Two topics in particular were repeatedly brought up: LLMs and Multi-Modal Models. Everyone at the GenAI Summit that I spoke with had some experience dealing with either digital automation and AI agents, but one session in particular caught my attention.
One of the panel discussions at the conference covered how Large-Language Models (LLMs) have scaled heavily in the past 5-6 years, their use is in fact only on the rise. Two of the panelists, Dr. Devandra Chaplotand Sharan Narang (research engineer at Meta AI), spoke in depth about the relationships between pre-training, synthetic data, and fine-tuning for more efficient processes.
Dr. Chaplot also discussed the scalability of multi-modal models given their current state. He explained that despite the cost of computing, multi-modal models were only at the beginning of their vast set of capabilities. As engineers work to improve the models over time, the gap between small and large models would eventually reduce as they became more powerful and cost-effective.
This was the first year of the GenAI Summit in San Francisco, but it was also the first event I was able to attend as an Ushur marketing intern. It was a week full of interesting conversations and conference attendants, and I appreciated getting to hear feedback on Ushur’s point of view on Generative AI for regulated enterprises. I see overlap in how CaseText used AI Agents to replicate legal assistants’ work, and how Ushur uses AI Agents to enhance customer experience professionals’ work. I also believe adoption of Generative AI is only going up from here as excitement from events like this Summit continue to grow. I plan on attending next year's GenAI Summit, too.
