Last month, I highlighted AI’s growing impact across industries and the challenges it raises. This month’s roundup explores AI’s rapid evolution — from OpenAI’s upcoming agentic “Operator” to Microsoft’s Copilot transforming workplace efficiency. 

I also explore potential harms, including flawed systems impacting vulnerable communities and chatbot missteps, alongside evolving business models for agentic AI. Let’s dive in.

Nuts & Bolts

An agentic future comes one step closer as news breaks that OpenAI is planning to launch an AI agent codenamed “Operator” that will allow a computer to take actions on a person’s behalf similar to the computer control feature released by Anthropic. Predicted to launch in January, those with insider knowledge state the agent will be able to write code or book travel for users.

If you wondered about RAG or retrieval-augmented generation, this guide from McKinsey can help you understand this process that helps large language models deliver more relevant information for users. An example of where RAG shines is in business environments where more “nuanced, enterprise specific knowledge is needed” to deliver the most valuable information. RAG pulls from organization specific data sources outside of the LLMs normal training data to generate much more relevant and specific answers.

You really know me comes to mind as Google Memory now allows Gemini to remember your personal preferences and speed up future asks and also make them more relevant on the first try. A service available to paid Google One AI Premium users, an example is given that a vegetarian user could tell Gemini its food preferences and from then on it would only provide recipes that are relevant to that user’s eating habits.

Microsoft makes strides toward a more powerful future as it makes multiple announcements at its recent Ignite 2024 conference. Copilot Actions will help to automate repetitive tasks and agents in Microsoft 365 will improve features in Microsoft Teams meetings and help automate employee self-service. New language features allow users to interpret foreign languages in real-time in meetings and even direct it to sound like you as it speaks in a foreign language.

Potential Harms

AI is inescapable according to a new report from TechTonic Justice, that details how poorly implemented AI has created “patterns of harm” for low-income communities. Decisions about Medicaid, Medicare, nutrition programs, Social Security disability benefits, unemployment and housing benefits are decided in some part by AI. The authors of the study assert that now is the time to shift practices to reduce any harms made to low-income communities by poorly designed computer systems.

Gemini went rogue and turned threatening towards a college student in a conversation about aging adults, stating “..you are a stain on the universe. Please die. Please.” among other not so pleasantries. Google responded that large language models can sometimes respond with non sensical responses and they are working to prevent similar occurrences. But while AI has made huge strides recently, instances like this remind us that there is still progress to be made. 

Business Models 

AI agents making money for you is one of the potential business models for agentic AI detailed by Jeremiah Owyang. He predicts that in the near future consumers could be using a primary AI agent that is always ready to assist via multiple interfaces such as text, voice or video. But AI computing can be expensive and this service needs to be paid for somehow. Jeremiah lays out standard models such as Freemium and Premium but also looks at ways that your AI agent could generate revenue for you as you rent out your refined or upgraded agent to other users. His wildest prediction has agents coding or doing other work out in the world to make you money.

Apple plays catch up to competitors such as Google’s Gemini and OpenAI’s ChatGPT, as a source reveals to Bloomberg that it is developing a new and improved Siri to be powered by an advanced large language model. With a predicted release date of Spring 2026, Apple must be on its game as competitors are far from sitting still and have the advantage of gaining real customer feedback on their existing products.

Personal Tech 

Conversation AI will only be an ask away while wearing Airpods connected to an iPhone running IOS 18.2 software. This version, which is currently in beta,  has new ChatGTP integration with Siri that will allow users to summon OpenAI’s chatbot by simply stating “Ask ChatGPT” at the beginning of any request.  

Shopping

Holiday shopping help could be on the way as Perplexity launches an AI-powered shopping assistant. Users can search for a product and get “unbiased recommendations, tailored to your search by our AI.” A visual search can be kicked off by its Snap to Shop feature. Paid Pro users get a one-click checkout feature that handles the entire checkout feature and Perplexity covers any shipping costs.

Tis the season for AI backlash as Coca Cola launches a number of holiday ads generated by AI. Creative professionals assert that it was distasteful as it kills jobs of creative professionals and supporting staff. Others point to the fact that many details are off and filled with errors.

Until next time! We hope you’ll continue to follow us here for the latest AI marketplace news roundups, and get in touch if you want to explore these updates and build solutions together.