This week, OpenAI and Google made major announcements about their new AI products. While they are both leaders in the AI space, they are each taking their products in very different directions, with significant impacts to the business world. Let’s explore them.
OpenAI Has Your New Personal Assistant
On Monday, OpenAI unveiled their newest AI model, called GPT-4o. The “4” indicates that it operates at a similar level of capability to their existing GPT-4 model, which has now been out for over a year.
However, the “o” stands for “omni”, which is what makes this model special. Previous models offered speech input/output, but it was an add-on to the existing text model: speech was translated to text, processed through the model, then text outputs converted into speech again. The new “omni” model has audio, images, and text baked directly into a single model (in AI parlance, it’s “multimodal”). Since no conversion is necessary, it operates faster and delivers better results without all the extra work.
The benefits for business are three-fold:
- Along with being faster, the new model is now less expensive – half the price of the previous model. (Note: this new price only applies to the API; if you’re using ChatGPT Pro, Teams, or Enterprise, your price remains the same, but you get a better model!)
- With it’s enhanced speech and built-in multi-lingual capabilities, it’s now a powerful meeting assistant. In a recent post, I discussed Zoom’s AI Companion that summarizes meetings, takes notes, and more. OpenAI just offered a near drop-in replacement with GPT-4o.
- The speech output for their new model is very human-like – expressive, creative, empathetic. Where customer-facing chat was previously exclusively text, this offers the potential for customer service bots that sound like humans instead of, well, robotic.
OpenAI is also rolling out a desktop app that leverages GPT-4o for a higher level of interactivity. This app can see your monitor, potentially control your programs (with your permission), and makes the AI appear instantly with a key press. The app is now available for Apple products, with a Windows version expected later this year.
Google Workplace Offers New AI Teammates
Where OpenAI wants you to build your workflow around their systems, Google is taking the opposite approach, enhancing their existing apps with a variety of powerful new AI capabilities. The most important one comes from their new Google Workplace feature: AI Teammates.
While Google has been adding AI features within specific Google Workspace products for a while, AI Teammates exist as an AI between your Workspace files. It can incorporate information from multiple documents, group chat history, and other sources to do work on behalf of your team. It just requires a quick description when creating your “team member” to guide the work you want it to do. In their presentation, Google demonstrated an AI project manager that captured decisions and raised blockers – all with references back to the original information.
I won’t get into all of the ways this can be used, but it’s huge. If your company needs a role where the work is based on review or insight solely from developed project documents, this could be a force multiplier for team coordination and execution. This is the new AI-enhanced future GenAI promised – and, right now, it’s exclusively available in Google Workplace.
Google Leans Hard Into AI-Enhanced Education
Since high-performing GenAI has become available, educators have expressed equal parts excitement and concern about what this means for education. If GenAI can do the work, what does this mean for student learning?
Google is making two major moves that will transform education.
First, given the potential for GenAI to be personalized tutors for students, Google developed LearnLM, a multimodal model focused specifically on educational tasks. Built directly into search, YouTube, Android, and Gemini, it’s designed to not just directly give an answer, but instruct students on solving the problem step by step.
Second, they are building AI-enhanced tools for educators within Google Classroom, including tools to help with building lesson plans and generating questions for students. These will be great resources to help teachers spend less time on paperwork and more time helping students.
Google Explores AI Agents
One concept of GenAI systems is as an all-knowing oracle: ask it a question, it gives you a fully developed answer. Except, some answers require more than one step to solve, requiring thought, reflection, iteration. LLM’s may mimic this, but they don’t actually do this under the hood. This is why they tend to be poor at directly answering some questions, like math problems (although they’re great at writing code that can solve them).
The solution to this is to view GenAI not as an oracle, but as a block in a larger solution system. Rather than just one question/one answer, you may ask an LLM a question, give it’s output to another LLM, feed that output to a database, feed that result to another LLM, and so on. In this approach, the GenAI systems are referred to as “agents”, and the agents work together to solve a problem. Today, this approach is still in its infancy, but promises to dramatically increase the potential of Generative AI tools.
Google announced that they are developing an AI assistant which takes the first steps in that direction. Their “search” tool will now deliver an “AI Overview” to complex, multistep queries. The activities demonstrated included developing a meal plan for a student on a budget, or creating a detailed itinerary for a family that meets certain criteria and evolves over time.
Focusing on low-stakes challenges like this is a great place to start. Agentic workflows will be a very powerful tool for the future of AI in business, and mastering those will give Google an immense advantage going forward.
But What’s Available Today?
As with a lot of tech demos, many of the products shown are not yet available. For OpenAI’s offering, GPT-4o can be used now, although isn’t quite as full featured as the demo. With Google, many of the features are in pilot or early-access stage, but they expect to launch them fully later this year. When it becomes available, you’ll be sure to see more on AI for Business.
Want to stay informed on the most important AI news in just 30 minutes a month? Sign up for the Executive Summary newsletter by AI For Business.
If you liked this post, have future posts delivered straight to your inbox by subscribing to the AI For Business newsletter. Thank you!