AI Transformation Brief – November 2024

Author(s): Bill Wong

VOL 11 | NOVEMBER 2024

INFO-TECH AI INSIGHTS

AI Transformation Brief

Featuring AI best practices and insights to enable our members to strategize, plan, develop, deploy, manage, and govern AI-based technologies and solutions.

Dividing line for the headline section

In This Issue – AI in the News | AI Research Highlights | Vendor Spotlight | Upcoming Events & Resources

AI IN THE NEWS

Big Tech Vendors Adopt Nuclear Power for AI Compute Demands

Read Google's Nuclear Plans

Read Amazon's Nuclear Plans

Read Microsoft's Nuclear Plans

In October, major AI vendors announced plans to leverage nuclear power to address the growing demand for AI-powered applications.

On October 14, Google announced plans to purchase energy from small modular nuclear reactors (SMRs).
On October 16, Amazon announced an investment to establish SMRs.
On October 19, Microsoft announced its plans to restart the Three Mile Island nuclear power plant.

ANALYST ANALYSIS

Nuclear power has its advantages and disadvantages. Big Tech vendors are positioning nuclear power as a reliable source of clean energy, as it's a low-carbon source and can reduce greenhouse gas emissions.
Environmentalists draw attention to the fact that nuclear power plants produce nuclear waste (which requires safe storage for thousands of years), come with the risk of accidents, and are costly to build and operate.

In recent years, many countries have focused on SMRs, which represent the next generation of nuclear power plants. SMRs are designed to be modular and less expensive to build, and their smaller footprint allows more flexibility in terms of where they can be deployed. They produce low-carbon electricity and include advanced safety features to reduce the risk of accidents.

For better or worse, until a new way is discovered to train AI models with less data or energy, nuclear energy may become more common for the Big Tech vendors for the foreseeable future.

AI Agents – The Next Frontier in the Evolution of AI

Google Vertex AI Agent Builder

AWS Bedrock Agents

Microsoft Copilot Studio

Salesforce Agentforce

AI agents are autonomous systems designed to interact with their environment, perceive information, and take actions to achieve specific goals. They are software programs that can reason, learn, and make decisions independently. AI agents are:

Autonomous – capable of independent assessment, decision-making, and actions.
Goal-oriented – capable of performing tasks to achieve a specific objective.

Adaptable – capable of learning and adjusting their actions.

ANALYST ANALYSIS

The introduction of AI agents heralds a new era of AI-driven automation that will transform human-technology interaction. These autonomous systems are capable of learning and reasoning to optimize business outcomes and change the limits of what’s possible by automating many more kinds of tasks and personalizing the user experience so organizations can focus more on creative and strategic initiatives.

Key vendors have recently announced the availability of their AI agent development platforms, including Google Vertex AI Agent Builder, AWS Bedrock Agents, Microsoft Copilot Studio, and Salesforce Agentforce. AI agents will continue to evolve and proliferate, with potentially millions of agents being deployed in the very near future.

Google NotebookLM Goes Viral

Read the Google NotebookLM Announcement

On July 12, 2023, Google released NotebookLM, a virtual research assistant targeting researchers, students, and knowledge workers dealing with large volumes of text-based information. Users can upload documents (up to 50 in the current release) to NotebookLM, and the application can provide a summary of all the documents, answer questions on the content uploaded, and generate content to enhance the user’s understanding and use of the uploaded content.

Accuracy of the responses is improved since they are grounded in the data and documents uploaded.

ANALYST ANALYSIS

During this past summer, access to NotebookLM expanded globally and multimodal capabilities were introduced by leveraging Google’s Gemini 1.5 multimodal AI model. New features allow the application to also ingest images, audio, and video and summarize this data to improve a user’s ability to fact-check content, create study guides and briefing documents, and now create “Audio Overviews” (made available September 2024).

What makes the Audio Overview so engaging is that it can produce a podcast that features two personas conversing about the content in question and includes human-like banter, questions, and enthusiasm. The adoption of this tool has surged and seems to have caught the industry – including Google itself – off guard.

Having tested NotebookLM with some of my upcoming research, I can begin to understand why users might find this solution engaging and compelling. It’s not a feature I thought my research needed, but I can see how it can help draw attention and highlight certain content. It’s still a relatively new product and people are asking for more control (additional personas) and features. The industry is clearly aware that this application has exploded in popularity; Meta has just announced its open-source version, NotebookLlama.

Anthropic Unveils Computer Use: Agent-Driven Computer Interaction

Read the Anthropic Computer Use announcement

Read the Anthropic Developing a computer use model

On October 22, 2024, Anthropic announced the availability of Computer Use. Leveraging its current AI model, Claude 3.5 Sonnet, Anthropic is the first to offer the development of AI agents capable of using computers autonomously by teaching them general computer skills.

Benefits include improved productivity, automation of routine tasks, improved user experience, and workflow optimization.

ANALYST ANALYSIS

This feature enables AI agents to intelligently and autonomously interact with computer systems in a similar way as humans do, including:

Navigating software interfaces (i.e. selecting objects on screen and performing cursor movements or mouse commands).
Searching, accessing, and automating data input when given a prompt.
Automating and performing tasks.

Other vendors of agent-based solutions will most likely replicate this feature for their offerings. This feature is new – and error-prone for now – yet it promises new ways in which we will interact with technology in the future.

OpenAI Introduces ChatGPT Search

Read the OpenAI ChatGPT search announcement

On October 31, 2024 OpenAI released ChatGPT search, an AI-augmented search engine.

Since the introduction of ChatGPT, conversational AI has been disrupting the entire software industry, including the search engine market. In December 2022, Perplexity was released, an AI search engine and one of the first products to challenge Google. Since then, the Big Tech vendors have started to release their solutions for AI search.

ChatGPT Plus and Team users will have access to ChatGPT search now. In the upcoming months, OpenAI plans to make this available to users using the free version of ChatGPT.

ANALYST ANALYSIS

Google is the overwhelming market leader in providing search but has had to augment its capabilities with generative AI to remain competitive.

To summarize some of the major players' activity with respect to AI search:

May 2024: Google release AI Overviews, augmenting its search capabilities with generative AI summarizing the search results.
July 2024: Microsoft enhances Bing with generative AI.
July 2024: OpenAI releases SearchGPT, an AI-powered search engine.
October 2024: Google expands access to AI Overviews to over 100 countries.
October 2024: OpenAI rebrands SearchGPT to ChatGPT search and integrates the search capabilities into ChatGPT.

Although OpenAI is attempting to take market share away from Google, ironically, this latest announcement may reduce market share from Microsoft Bing instead, since many users have adopted Bing recently only to access ChatGPT. Users testing AI search engines should verify results, as responses may be fabricated/hallucinated by the AI component of the AI search engine.