šŸ§© OpenAI Collaborates with Broadcom & TSMC To Make Their Own AI Chip

OpenAI teams up with industry leaders Broadcom and TSMC to develop its first in-house AI chip, aiming to secure critical semiconductor supply and reduce reliance on Nvidia.

Author

Jack Lajoie
October 30, 2024 ā€¢ read time ~ 8 minutes

The Daily Current āš”ļø

Welcome to the creatives, builders, pioneers, and thought leaders ever driving further into the liminal space.

OpenAI's strategic partnership with Broadcom and TSMC marks a significant shift as the company ventures into developing its own AI chips, aiming to alleviate reliance on Nvidia amid semiconductor shortages. Meanwhile, GitHub is enhancing its Copilot tool by integrating multiple AI models from companies like Anthropic, Google, and OpenAI, giving developers increased flexibility and choice, despite breaking exclusivity with Microsoft. Google reports writing 25% of its code with AI, while the emergence of deepfake detection startups underscores the growing importance of AI in cybersecurity.

 šŸ”Œ Plug Into These Headlines:

OpenAI is developing its own AI chip in partnership with Broadcom and TSMC to address the shortage of advanced semiconductors needed for training large language models. This move aims to secure a stable supply of AI chips and reduce dependency on Nvidia, the current market leader. The company has scaled back its initial plans to build its own chip factory, opting instead to leverage existing expertise in chip design and manufacturing.

  • OpenAI is developing its first AI chip in collaboration with Broadcom and TSMC, with testing expected to begin in 2025.

  • The company has abandoned plans to build its own chip factory due to high costs and complexity.

  • OpenAIā€™s chip development is driven by the shortage of Nvidiaā€™s H100 AI chips, which are crucial for training large language models.

  • The partnership with Broadcom and TSMC leverages existing expertise in chip design and manufacturing.

šŸ¦¾ OpenAIā€™s move reflects the growing trend of tech companies developing custom AI chips to secure their AI infrastructure. Access to compute is still the primary bottleneck they need to avoid.

GitHub is enhancing Copilot by integrating multiple AI models from various providers, allowing developers to choose the most suitable model for their coding tasks. This update, starting with Copilot Chat, introduces models from Anthropic, Google, and OpenAI, each offering unique strengths in code generation and understanding. Despite moving beyond exclusivity with Microsoft, the move reflects GitHubā€™s commitment to providing developers with flexible, powerful tools that adapt to their specific needs and preferences.

  • GitHub announced the integration of multiple AI models into Copilot at GitHub Universe 2024.

  • The update includes models from Anthropic (Claude 3.5 Sonnet), Google (Gemini 1.5 Pro), and OpenAI (o1-preview and o1-mini).

  • Multi-model functionality will be rolled out progressively, starting with Copilot Chat.

  • Developers can select their preferred model in VS Code or on GitHub.com, while organizations can control model availability.

  • This update reflects GitHubā€™s commitment to being an open developer platform and providing flexible AI-assisted development tools.

šŸ”€ GitHubā€™s multi-model approach for Copilot reflects the reality that no single AI model excels at everything, perhaps signaling that the industry is heading towards more specialized and adaptable AI tools for developers.

Alphabetā€™s third-quarter earnings report highlights the substantial financial benefits of its AI initiatives. Google Cloud revenues surged 35% year-over-year to $11.4 billion, driven by AI offerings that have attracted new enterprise customers and expanded existing relationships. CEO Sundar Pichai emphasized that over 25% of new code at Google is now AI-generated, reflecting the companyā€™s commitment to AI-driven efficiency across its operations and products.

  • Google Cloud revenues grew 35% year-over-year to $11.4 billion in Q3 2024, driven by AI offerings.

  • AI tools have led to 30% deeper product adoption among existing customers, as reported in the Q3 2024 earnings call.

  • Googleā€™s search business revenue increased by 12.3% year-over-year to $49.4 billion in Q3 2024.

  • Alphabetā€™s stock value grew by nearly 30% in 2024, surpassing the S&P 500ā€™s 23% gain.

šŸ“Š Googleā€™s financial gains and reported depth of AI integration support the theory that early adopters of AI technologies will see substantial returns on investment, likely justifying the exorbitant expenditures in the long run.

  • Users can quickly & easily bring up a chat to reference, or pick up a chat where they left off.

  • Plus and Team users will have access within the day. Enterprise and Edu users will have access in one week. Free users will start getting access throughout the next month.

Deepfake detection technology is emerging as a critical security tool for businesses due to the rapid advancement of AI-generated fake audio and video content. Investors are increasingly funding startups in this field, recognizing that these tools may soon become as essential as email and application security.

  • Norwest Venture Partners categorized the deepfake detection ecosystem into six main areas, including image and video detection, audio detection, and content watermarking.

  • A recent study found that 56% of participants believed a deepfake video was real, and 50% thought a deepfake audio clip was legitimate, highlighting the difficulty in distinguishing fake content.

  • Investors are increasingly funding deepfake detection startups, with Reality Defender raising $33 million and Clarity raising $16 million in recent rounds.

  • Pindrop, a leader in audio detection, has seen a significant increase in deepfake incidents among its customers, from one per month in 2023 to one per day per customer in 2024.

šŸ˜µā€šŸ’« The growing threat of deepfakes has evolved from a theoretical concern to a real and immediate problem for businesses, driving the need for effective detection solutions.