AI news highlights: Gemini Ultra, Cisco-Nvidia partnership

Google releases Gemini Ultra

Back in December, Google announced Gemini, its most advanced model to date. The multimodal Gemini model powers the Bard chatbot, can handle complex and domain-specific questions and offers significantly improved speed and accuracy on coding tasks.

Google has made it clear that Gemini is more than just another model: It represents a mindset shift. Gemini supports the foundation of an entire Google AI ecosystem, from products and assistants targeting end users to APIs and platforms for builders and developers.

Back in December, only the Pro tier of Gemini was available through Bard. While it was more sophisticated than other Google models, it performed at around the same level as OpenAI's GPT-3.5. But last week, Google released its Ultra model to users through a new AI assistant called Gemini Advanced. Google says that Gemini can outperform human experts on massive multitask language understanding, a method that uses a combination of 57 subjects -- including math, physics, history, law, medicine and ethics -- to test knowledge and problem-solving abilities.

The other difference between Gemini Pro and Gemini Advanced, which uses the Ultra version of the model, is cost. Gemini Advanced requires a subscription to the new Google One AI Premium plan.

Deloitte bolsters AI practice with OpTeamizer acquisition

Research from TechTarget's Enterprise Strategy Group (ESG) has shown that the top challenge organizations today face when it comes to implementing generative AI is a lack of employee expertise and skills. In light of this skills gap, the natural next question is where organizations are turning for help -- and ESG's research showed that the answer is management consultancies.

On Feb. 5, Deloitte announced its acquisition of OpTeamizer, which specializes in building and implementing AI that runs on Nvidia-accelerated computing hardware. This is great news for Deloitte customers, who are already seeing value from Deloitte's existing alliance with Nvidia. OpTeamizer's AI and data science experts have extensive experience with Nvidia technology, including everything from training workshops to high-performance computing and Nvidia's CUDA software framework, just to name a few.

This acquisition presents a massive opportunity for Deloitte to expand its portfolio, capabilities and generative AI presence across the entire AI stack. Additionally, it highlights Deloitte's commitment to its customer-facing generative AI initiatives.

Cisco and Nvidia extend partnership on AI deployment and management

At the Cisco Live conference in Amsterdam on Feb. 6, Cisco and Nvidia announced an expansion of their partnership that extends into the data center, with integrated data center offerings. This collaboration includes the integration of Nvidia's Tensor Core GPUs into Cisco's M7 UCS rack and blade servers, jointly validated reference architectures and support for the Cisco Networking Cloud. It also encompasses digital experience monitoring using ThousandEyes, Cisco Observability Platform and more.

Vendors are increasingly putting together full-stack offerings with their partner ecosystems, and this announcement continues the trend of simplifying and accelerating time to value. The reference architectures, in particular, highlight this, combining Cisco and Nvidia's capabilities with partners such as Pure Storage, NetApp and Red Hat to take the guesswork out of deploying AI infrastructure.

What might get overlooked in this announcement is on the networking side. Obviously, Cisco is a leader in networking, but the more interesting aspect here is Nvidia's involvement, given the company's interest in InfiniBand to support AI. Although Nvidia will continue to push InfiniBand, this announcement highlights Nvidia's recognition of the need to support customers' networking preferences.

Hugging Face to challenge OpenAI on AI assistant front

I would argue the easiest and most natural generative AI use case today is simply using an AI assistant, which can streamline operations, enhance efficiency and improve productivity. The value proposition of these assistants is their ability to automate routine tasks, reduce errors and tirelessly process information.

Although there are several enterprise generative AI assistants on the market -- including Microsoft's Copilot, Google's Duet AI and AWS' Amazon Q -- organizations continue to explore ways to build their own.

OpenAI's custom GPT builder enabled this functionality, but it requires a paid subscription and only works with OpenAI's proprietary LLM, ChatGPT. Many organizations want to build similar functionalities using their preferred open source models.

Hugging Face has made this possible through the launch of third-party, customizable Hugging Chat Assistants, Hugging Face's alternative to OpenAI's GPTs announced on Feb. 2. With Hugging Chat Assistants, users can create custom versions of Hugging Face Chat in as few as two clicks. What's more, they can do so for free, using whichever open source LLM they'd like to power the AI assistant.

IBM announces new AI Alliance members and working groups

There is such a robust ecosystem of vendors, partners and consultancies driving the adoption of trusted AI. It's a big reason why IBM, Meta and many supporting vendors created the AI Alliance, a group of large tech vendors, startups, academics and others who see the value of an open, collaborative ecosystem and pledge to contribute to furthering those same goals. The alliance's focus is on accelerating and spreading open innovation across the AI technology landscape to improve foundational capabilities, safety, security and trust in AI, and to responsibly maximize benefits to people and society everywhere.

I was disappointed to see some big vendors missing from the membership list, including Google, Microsoft, AWS, Nvidia and OpenAI. The good news is that new members continue to pile in -- and while it's not anyone on the aforementioned list, there are still some extremely big names in tech, including Databricks, Snowflake and Uber.

In my view, the biggest component of this announcement is the two new working groups: the AI Safety and Trust Tooling group, and the AI Policy Advocacy working group. The goal of these groups is to bring together leading researchers, developers, policymakers and industry experts to address the challenges of generative AI and democratize its benefits. This includes everything from outlining best practices on AI safety, trust, ethics and cybersecurity, to establishing a definitive set of benchmarking capabilities, to publicly sharing information on key policy topics, including red teaming, regulation of applications and access to hardware.

Mike Leone is a principal analyst at TechTarget's Enterprise Strategy Group, where he covers data, analytics and AI.

Enterprise Strategy Group is a division of TechTarget. Its analysts have business relationships with technology vendors.