CONVO OVER COFFEE ☕

Big topics to discuss with friends and colleagues

Google's Gemini: A Groundbreaking Multimodal AI Model 🚩
Google has recently announced the release of Gemini, a new family of large language models (LLMs) that marks a significant leap in AI capabilities. Gemini is touted as Google's most ambitious and advanced AI project yet. This model is set to transform technology interactions and open up new possibilities across various industries. Gemini is introduced in three versions tailored for distinct applications:

• Gemini Ultra: The most powerful variant, excelling in complex reasoning tasks and surpassing human-level performance on several benchmarks.

• Gemini Pro: Balanced for scale and speed, it's ideal for commercial applications.

• Gemini Nano: Designed for efficiency, this model can operate on smartphones and other portable devices.

My take: Gemini Ultra has achieved unprecedented results, outperforming state-of-the-art outcomes on 30 of 32 key AI research benchmarks and even surpassing human scores in broad knowledge and problem-solving tests. Unlike previous models limited to text, Gemini is inherently multimodal, comprehending and integrating text, images, audio, video, and other formats. What is core here is Gemini is the first model to actually rival top OpenAI models. It took Google a year to catch-up. A year! But this means one thing - now we will see even more competition.

IBM and Meta Launch the AI Alliance for Open and Responsible AI Development 💚
IBM and Meta have spearheaded the formation of the AI Alliance, an international consortium aimed at propelling open innovation in artificial intelligence. This global community is described as a collaborative effort involving over 50 leading entities from various sectors committed to advancing AI in a manner that emphasizes trust, transparency, diversity, and responsibility. Notable Absences? Despite its impressive roster, leading AI research labs and model providers like OpenAI, Anthropic, Google, Microsoft, NVIDIA, Inflection, Cohere, Amazon, and others are not listed as participants.

My take: The success of such a consortium will largely depend on its ability to effectively coordinate diverse interests and translate its ambitious goals into tangible outcomes. The absence of some key AI players might also pose challenges in achieving comprehensive industry-wide collaboration. Nonetheless, it is a sign of polarization in the AI world - more commerce vs more openness. Meta being the most vocal one for the open-source is not a surprise for me. Fact that in such Alliance we do not see OpenAI - which as you most likely guessed right - has Open in it’s name not for domain availability reasons.

Runway and Getty Images Unite to Offer Commercially Safe AI Video Tool for Enterprises 💚
Runway, a leading AI video startup, is teaming up with Getty Images to introduce a new AI-powered text-to-video model catering to the commercial sector. This partnership is portrayed as a significant advancement for businesses, offering them an innovative tool for video content creation with enhanced legal safeguards against copyright infringement. This is an exclusive collaboration: Runway is set to release a custom text-to-video model, trained using Getty Images' licensed image and footage library, specifically tailored for enterprise-level commercial use.

My take: By leveraging Getty Images' extensive licensed content, RGM (model that both parties will create) will addresses a critical pain point in AI-generated media: copyright compliance. This collaboration not only promises to streamline video production for businesses but also sets a precedent for responsible AI usage in content creation. I think we will see a lot more of these partnerships going onward for companies unwilling to partner with giants like Google or Microsoft. The reason? Access to data.

Other news:

• Elon tries to raise $1 billion for X.ai 🚩

• Meta does not stop at AI Alliance and launches Purple Llama 💚

• Hardware wars intensify with AMD offering MI300 for AI and HPC 🟠

• Bard gets an update with Gemini Pro with Mark Rober showing how! 🟠

• Microsoft Bing takes DeepSearch onboard in power move against Google 🟠

• Hype goes up - Mistral AI in talks to secure €450M at $2B valuation 🚩