SAN FRANCISCO (AP) — Less than a year into its meteoric rise, the company behind ChatGPT unveiled the future it has in mind for its artificial intelligence technology on Monday, launching a new line of chatbot products that can be customized to a variety of tasks.
OpenAI CEO Sam Altman addressed a crowd of over 900 software developers and attendees at their first developer conference. He expressed that in the future, users will be able to simply request what they need from a computer and it will handle all necessary tasks. This conference follows the tradition of technology exhibitions, popularized by Apple in Silicon Valley years ago.
During a gathering at a spacious former Honda dealership in San Francisco, OpenAI introduced a latest edition known as GPT-4 Turbo, which boasts enhanced capabilities and the ability to access details on global and cultural occurrences up to April 2023. This is a significant improvement from previous versions, which were unable to provide answers about events beyond 2021.
The company has introduced a new iteration of its AI model, GPT-4V, which incorporates vision capabilities. This allows the chatbot to analyze images. In a recent research paper published in September, the company demonstrated how this tool can provide descriptions of images to individuals who are visually impaired.
According to Altman, ChatGPT has over 100 million users who are active on a weekly basis and 2 million developers, all of whom were reached through word-of-mouth.
He introduced a fresh series of items named GPTs – with a focus on the plural form – which will allow individuals to create personalized versions of ChatGPT for particular duties.
Alyssa Hwang, a computer science scholar at the University of Pennsylvania, was given an early preview of the GPT vision tool. She praised its ability to accurately describe a variety of complex images, but also noted that there is room for improvement.
An example of this is when Hwang added a picture of steak and wrote a comment about chicken noodle soup, causing the chatbot to incorrectly describe the image as being related to chicken noodle soup.
Hwang warned that this could potentially result in adversarial attacks. For example, if someone were to embed offensive text into an image, they could receive unwanted results.
This is one reason why OpenAI has granted early access to researchers like Hwang to identify any potential issues in their latest technologies before they are widely released. Altman explained on Monday that the company takes a gradual and iterative approach to deployment, allowing for time to address any safety concerns.
The road to OpenAI’s first DevDay has been an unconventional one. Initially established as a non-profit organization for research in 2015, it gained global recognition less than a year ago when it unveiled a chatbot that has sparked both enthusiasm and concern, leading to calls for international regulations to steer the rapid progress of AI.
The conference takes place one week after President Joe Biden’s signing of an executive order that will establish initial guidelines for AI technology in the United States.
Under the Defense Production Act, the mandate stipulates that AI developers, including OpenAI, Microsoft (its financial supporter), and competitors like Google and Meta, must disclose details to the government regarding AI systems that have the potential to reach “high levels of performance” and potentially pose safety hazards.
The arrangement was based on self-imposed pledges made by prominent AI creators at the beginning of this year, as directed by the White House.
There is a great deal of anticipation surrounding the economic potential of the newest generation of generative AI tools. These tools have the ability to generate text, images, sounds, and other media based on written or spoken prompts.
Microsoft CEO Satya Nadella briefly joined Altman on stage, expressing his love for the audience amidst their cheers.
Nadella highlighted Microsoft’s role as a business ally by providing data centers to OpenAI for enhanced computing capabilities to develop more sophisticated models.
“I am confident that our partnership in the tech industry is unparalleled. I am eagerly looking forward to collaborating on the development of AGI,” stated Altman, alluding to his objective of creating artificial general intelligence that can match, or even surpass, human capabilities in various tasks.
Several chatbot companies, such as Bing by Microsoft, use OpenAI’s technology. However, there are now more competitors emerging, such as Google’s Bard and Anthropic’s Claude, a startup in San Francisco founded by ex-OpenAI employees. OpenAI also has competition from developers who make their code and system components available for free through open source models.
The latest rival of ChatGPT is Grok, which was introduced by billionaire CEO of Tesla, Elon Musk, on his social media platform X (previously known as Twitter) over the weekend. Musk, who was a co-founder of OpenAI and has since left the company, has now launched a new venture called xAI to make his own impact on the speed of AI advancement.
Grok is currently only accessible to a select group of initial users, but guarantees to provide responses to challenging questions that other chatbots refuse to answer due to protective measures in place to prevent offensive replies.
When a reporter asked for a comment on the timing of Grok’s release, Altman responded with “Elon’s gonna Elon.”
According to Gartner analyst Arun Chandrasekaran, a significant portion of OpenAI’s announcement on Monday was focused on addressing the concerns of businesses interested in incorporating ChatGPT-like technology into their operations.
“Offering more affordable products was a major request, along with the ability to personalize AI models using an organization’s internal data,” explained Chandrasekaran. He also mentioned that the “Copyright Shield,” where OpenAI covers the expenses of defending customers from copyright infringement cases related to the training of models on large amounts of online content, was another attractive aspect for businesses.
According to Goldman Sachs’ forecast from last month, the implementation of generative AI could potentially increase labor productivity and result in a 10% to 15% long-term growth in the global gross domestic product, which measures the overall production of goods and services within an economy.
In his description, Altman envisioned a world where AI agents could assist individuals with a wide range of tasks in both their professional and personal lives.
The speaker stated that individuals desire AI that is more intelligent, tailored to their needs, and can accomplish more tasks on their behalf.
O’Brien reported from the city of Providence in the state of Rhode Island.
The Associated Press and OpenAI have reached a licensing agreement that permits the use of a portion of AP’s written records to train the tech company’s extensive language model. AP is compensated with an undisclosed fee for the utilization of its material.