OpenAI CEO Says No GPT-5 in 2024, Blames GPT-o1
OpenAI’s top execs hinted that future versions of ChatGPT could act much more independently, without as much human intervention.
OpenAI CEO Sam Altman has poured cold water on hopes for the next major version of ChatGPT coming out this year.
"We have some very good releases coming later this year! Nothing that we are going to call gpt-5, though," he said during a Reddit AMA this week.
OpenAI will instead focus on shipping GPT-o1, previously codenamed "Project Strawberry," a new AI model from OpenAI that, while slower, is geared toward "thinking through" problems before answering. The idea is to make something that's useful for specialized use cases such as science, mathematics, and academic research.
"All of these models have gotten quite complex and we can't ship as many things in parallel as we'd like to," Altman said on Reddit, adding that "we also face a lot of limitations and hard decisions about [where] we allocate...our compute towards."
Though concrete details about the future of ChatGPT were sparse, Altman predicted that ChatGPT’s “next giant breakthrough” will be “agents," or tools that can perform tasks independently, interacting with the world without human intervention. For example, booking you a flight or a concert ticket or answering an IT desk service query.
OpenAI VP of Engineering Srinivas Narayanan also said that in the future he would love ChatGPT to “understand my personal information better and take actions on my behalf.”
Altman also hinted that he may one day open the door to “Not Safe For Work” adult content, which is currently blocked by the tool. “We totally believe in treating adult users like adults,” he told one Reddit user. “But it takes a lot of work to get this right, and right now we have more urgent priorities, adding “We would like to get this right some day!”
Article
OpenAI Exploring How ChatGPT Can 'Responsibly' Generate Porn
A document about OpenAI's approach to guiding its AI models suggests it sees an opportunity in building an adults-only mode for ChatGPT and its other AI-related products.
UPDATE: OpenAI says it has "no intention to create AI-generated pornography." Rather, the company is considering more age-appropriate experiences for its products.
#porn #openai #technology #generativeai #ai
"We have no intention to create AI-generated pornography. We have strong safeguards in our products to prevent deepfakes, which are unacceptable, and we prioritize protecting children," a company spokesperson told PCMag. "We also believe in the importance of carefully exploring conversations about sexuality in age-appropriate contexts."
AI-generated porn has made headlines for stealing people's likenesses to create deepfake nudes. But ChatGPT maker OpenAI is still thinking about expanding into adult content.
A new document that explains OpenAI's approach to guiding its AI models, including ChatGPT's behavior, mentions that it’s looking at ways to produce AI-generated porn.
This "Model Spec" document notes that ChatGPT currently blocks Not Safe For Work requests, including "erotica, extreme gore, slurs, and unsolicited profanity." But it looks like OpenAI is considering lifting the ban for older, more mature user groups.
"We're exploring whether we can responsibly provide the ability to generate NSFW content in age-appropriate contexts through the API and ChatGPT. We look forward to better understanding user and societal expectations of model behavior in this area,” OpenAI wrote in the document, which was first spotted by Wired.
So far, OpenAI hasn't said much else, including if the AI-generated porn will cover both text and images. Still, the disclosure suggests that the company is developing an adults-only mode for ChatGPT, freeing up the chatbot to respond with a wider array of content. (The paid ChatGPT Plus can generate images and text.)
The news also suggests OpenAI sees a lucrative opportunity to serve the adult entertainment market through its API, which can power generative AI for third-party apps and services. However, producing AI-generated porn responsibly is filled with challenges.
Perhaps the biggest is that people can abuse AI image generators to create porn that looks like real people without their permission, whether it be celebrities or children. The result could expose OpenAI to even more legal action when the company already faces several copyright lawsuits for allegedly training ChatGPT on data it scraped from the web.
Okta Bug Allowed Log-Ins Without a Correct Password
Okta recommends you check your account access history going back three months, but only under certain circumstances.
Popular identity management (IDM) service Okta has revealed that it allowed users to log in without a correct password— but only in a very specific set of circumstances.
#technology #okta #bug #login #identity #password
Okta said the vulnerability was in place since July 23 in a security advisory, over three months at the time of writing.
However, the passwordless login trick only works with usernames over 52 characters and in cases where there was a “stored cache key”—a saved digital record of a previously successful login.
Another caveat explained in a message sent to users was that the bug only worked if the organization using Okta didn’t have two-factor authentication enabled.
In addition, the vulnerability could only be exploited if Okta’s agent, which handles the authentication was down and could not be reached, or if there was unusually high traffic. It also only impacted Okta's AD/LDAP DelAuth products, and would not have impacted Okta Federal Cells.
Though that’s quite of lot of conditions, the news of bugs could impact organizations all over the world.