In the spirit of advancing innovation in AI, we recently hosted a panel discussion on AI deployment led by Jack Erickson, Product Marketing Manager at Intel. He was joined by Intel® Liftoff members, Meryem Arik (Co-founder of TitanML) and Alex Porter (Co-founder of Mod Tech Labs), as well as Peter Ma (Co-founder of SiteMana) and Ramtin Davanlou (CTO - Accenture & Intel Partnership). It was a chance to reflect on some of the challenges that AI startups are facing, and the role of Intel® hardware and software in breaking through these barriers to achieve the vision of “AI everywhere”.
From Research to Production: Key Challenges & Solutions
The panel began by discussing the biggest barriers to AI deployment. Cost featured as a major obstacle in various forms, including access to appropriate hardware and GPU infrastructure. The margin on GPUs was discussed extensively, especially its limiting effect on scalability.
Panellists had differing views on how to tackle the problem of cost. One proposed solution was to simply wait for the price of GPUs to come down over time. On the other hand, Meryem raised possibilities like pruning, compression and reducing memory to optimize LLMs. She also pointed to the significant reduction in inference cost that TitanML’s Takeoff provides for ML teams. The panel also agreed that Intel® technology plays a crucial role in making AI applications faster and more scalable, particularly Intel® Xeon® and Neural Compressor.
On another note entirely, Alex raised the issue of customer education. In this arena, there is an ongoing debate about the role of AI. Her startup, Mod Tech Labs, has to spend a lot of time educating customers to overcome reservations about its impact on the industry and the professionals in it. In addition, AI needs to be deployed in a way that is non-technical for the end user.
Featured founders
Alex Porter is an innovator in AI-powered real-time content and Intel® Top Innovator. She is also CEO of Mod Tech Labs, an AI startup that provides 3D automation tools for real-time content. Mod Tech Labs aims to empower artists and content creators through their vision to “make 3D easy”.
Alex's leadership at Mod Tech Labs has clearly delineated a trajectory toward simplifying complex 3D content creation. This commitment to innovation found a strong ally in the Intel Liftoff program, as Porter explains:
With Liftoff's assistance, we were able to increase our speed by 32x and doubled our precision for anomaly detection in media content.
We see this partnership with Intel Liftoff as a validation of our vision and potential. We look forward to continuing our journey with their support and are excited about the opportunities that lie ahead.”
TitanML's Titan Takeoff Inference Server streamlines LLM deployments, delivering 10x speed improvements and 90% cost reductions, along with major decreases in latency. It employs hardware-aware acceleration techniques to make models faster, even on CPU. TitanML also ensures privacy, making AI deployment not only faster but safer.
Overcoming Barriers to AI with Intel® Liftoff
The session turned out to be a brief but highly informative discussion of the biggest barriers to AI deployment, as well as some of the most promising routes around them.
We invite you to explore Intel® Liftoff for Startups, a program that empowers emerging AI ventures to unlock their full potential and achieve the vision of "AI everywhere." Apply today to start your journey or get in touch with our program team at intel.liftoff@intel.com.
I'm a proud team member of the Intel® Liftoff for Startups, an innovative, free virtual program dedicated to accelerating the growth of early-stage AI startups.