OpenAI’s GPT-4 is Here, Alongside More Breakthrough Gen AI Advancements

Generative AI is back in the headlines after Google, Microsoft, OpenAI, and Salesforce each announced new advancements and integrations this week that are poised to further revolutionize the technology landscape. Four months after ChatGPT took the world by storm, tech giants are racing to integrate the wonders of generative AI into seemingly every product offering, with many of this week’s advancements focused on making work more efficient. As noted by analysts, even the timing of such announcements—one after another—demonstrates the attempted one-upmanship and a “point of competition” between tech rivals as the AI arms race heats up. 

Key Takeaways

OpenAI released the highly-anticipated GPT-4 this week, just four months after ChatGPT led to a fundamental shift within the tech industry. GPT-4 is already available via a new version of ChatGPT and alternative AI chatbots. 

Recently, both Google and Microsoft unveiled new integrations and ways of working with generative AI across Google Workspace and Microsoft Office tools that will be launching soon.

Every day brings more startups and industry leaders announcing generative AI and large language model (LLMs) integrations, with the latest being Salesforce and new plans to offer access to ChatGPT through Slack. 

OpenAI’s next-generation AI language model was introduced earlier this week and is now available via waitlist and through some alternative tools like Microsoft’s AI-powered Bing search engine. Some of the most significant differences between GPT-4 and its predecessors include: 

GPT-4 is multimodal, meaning it can generate content based on text and images. 

The new GPT-4 can interpret and generate longer blocks of information. 

GPT-4 demonstrates “human-level performance” on various academic and professional certification tests, receiving high marks on the LSAT, GRE, SATs, AP exams, and a simulated law school bar exam, according to OpenAI.

Improved problem-solving abilities, allowing GPT-4 to handle more complex and nuanced problems and queries. 

However, according to OpenAI, “In a casual conversation, the distinction between GPT-3.5 and GPT-4 can be subtle. The difference comes out when the complexity of the task reaches a sufficient threshold—GPT-4 is more reliable, creative, and able to handle much more nuanced instructions than GPT-3.5.” As for the trustworthiness of outputs by GPT-4, OpenAI noted that the technology has similar limitations to earlier GPT models and is still not “fully reliable” as it can be found to ‘hallucinate’ facts and make errors. 

Related: Read PMG’s POV on generative AI here.

The announcement of GPT-4 was met with mixed reception from the AI community. Much to the surprise of experts and researchers, GPT-4 is not an open AI model. Critics argued that despite the 98-page research paper introducing GPT-4, it contains little information about the data used “to train the system, its energy costs, or the specific hardware or methods used to create it,” according to The Verge. OpenAI defended the decision to release GPT-4 as a closed model, with OpenAI’s chief scientist and co-founder Ilya Sutskever explaining in an interview with The Verge that its reasons were “self-evident” as the generative AI landscape matures and competition heightens. Analysts note that the launch of GPT-4 was only a few weeks after Meta’s new AI language model, named LLaMA, leaked online, sparking “similar discussions about the threats and benefits of open-source research.” 

Google announced new AI advancements on Tuesday that are designed for businesses, creators, and developers across Google Workspace, Google Cloud, and a new intuitive prototyping environment that’s known as MakerSuite. Google is also introducing PaLM API, an easy, secure way for developers who are experimenting with AI to build apps and technology on top of Google’s large language models (LLMs). Accompanying PaLM API is the MakerSuite, which will help users “quickly prototype ideas, and over time, will [include] features for prompt engineering, synthetic data generation, and custom-model tuning,” according to the Google announcement. 

Similarly, a set of new AI features are coming soon to Google Workspace, helping users draft emails, job descriptions, and other types of content directly within Google Docs, Gmail, and Google’s other workplace apps. 

In the coming months, Google Workspace will introduce a variety of AI capabilities enabling users to: 

Draft, reply, summarize, and prioritize content in Gmail

Create auto-generated images, audio, and video in Google Slides

Capture meeting notes in Google Meet

Enable new workflows in Chat and Google Sheets

Google Cloud CEO Thomas Kurain said in the blog post announcement that he expects this wave of AI advancements will have a “profound effect on every industry.” These AI-enabled tools and new features are available to select testers and developers now and will be rolling out more broadly in the weeks and months ahead.

Related: GM wants to bring ChatGPT-like assistant to drivers.

Microsoft hosted a customer event on Thursday to introduce Copilot, a new way to harness the power of generative AI directly within Microsoft workplace apps that’s coming soon to Excel, PowerPoint, Outlook, Teams, and Microsoft Word. Copilot promises to help generate emails, draft blog posts, summarize meeting notes, and more. At the event, Microsoft CEO Satya Nadella remarked that Copilot marks a new era of computing that stands to transform the way people collaborate and work. “As we look ahead, we believe this next generation of AI will unlock a new wave of productivity growth,” said Nadella. He later shared that these features would “remove the drudgery” from daily tasks and responsibilities. 

In the coming months, Microsoft will integrate OpenAI’s GPT-4 AI model to power Copilot across Microsoft’s suite of workplace productivity tools. While Copilot uses OpenAI’s underlying technology, Microsoft confirmed that the new features and integrations were built using Microsoft’s “comprehensive approach to security, compliance, privacy, and responsible AI.”