Become a fan of Slashdot on Facebook

 


Forgot your password?
Close
typodupeerror
×
AI Twitter

OpenAI Holds Surprise Livestream to Announce Multi-Step 'Deep Research' Capability (indiatimes.com) 56

Posted by EditorDavid from the X-marks-the-spot dept.
Just three hours ago, OpenAI made a surprise announcement to their 3.9 million followers on X.com. "Live from Tokyo," they'd be livestreaming... something. Their description of the event was just two words.

"Deep Research"

UPDATE: The stream has begun, and it's about OpenAI's next "agent-ic offering". ("OpenAI cares about agents because we believe they're going to transform knowlege work...")

"We're introducing a capability called Deep Research... a model that does multi-step research. It discovers content, it synthesizes content, and it reasons about this content." It even asks "clarifying" questions to your prompt to make sure its multi-step research stays on track. Deep Research will be launching in ChatGPT Pro later today, rolling out into other OpenAI products...

And OpenAI's site now has an "Introducing Deep Research" page. Its official description? "An agent that uses reasoning to synthesize large amounts of online information and complete multi-step research tasks for you. Available to Pro users today, Plus and Team next."

Before the livestream began, X.com users shared their reactions to the coming announcement:

"It's like DeepSeek, but cleaner"
"Deep do do if things don't work out"
"Live from Tokyo? Hope this research includes the secret to waking up early!"
"Stop trying, we don't trust u"

But one X.com user had presciently pointed out OpenAI has used the phrase "deep research" before. In July of 2024, Reuters reported on internal documentation (confirmed with "a person familiar with the matter") code-named "Strawberry" which suggested OpenAI was working on "human-like reasoning skills." How Strawberry works is a tightly kept secret even within OpenAI, the person said. The document describes a project that uses Strawberry models with the aim of enabling the company's AI to not just generate answers to queries but to plan ahead enough to navigate the internet autonomously and reliably to perform what OpenAI terms "deep research," according to the source. This is something that has eluded AI models to date, according to interviews with more than a dozen AI researchers.

Asked about Strawberry and the details reported in this story, an OpenAI company spokesperson said in a statement: "We want our AI models to see and understand the world more like we do. Continuous research into new AI capabilities is a common practice in the industry, with a shared belief that these systems will improve in reasoning over time." The spokesperson did not directly address questions about Strawberry.

The Strawberry project was formerly known as Q*, which Reuters reported last year was already seen inside the company as a breakthrough... OpenAI hopes the innovation will improve its AI models' reasoning capabilities dramatically, the person familiar with it said, adding that Strawberry involves a specialized way of processing an AI model after it has been pre-trained on very large datasets.

Researchers Reuters interviewed say that reasoning is key to AI achieving human or super-human-level intelligence... OpenAI CEO Sam Altman said earlier this year that in AI "the most important areas of progress will be around reasoning ability.
This discussion has been archived. No new comments can be posted.

OpenAI Holds Surprise Livestream to Announce Multi-Step 'Deep Research' Capability More

OpenAI Holds Surprise Livestream to Announce Multi-Step 'Deep Research' Capability

Comments Filter:

  • Investor Relations (Score:5, Interesting)

    by bill_mcgonigle ( 4333 ) * on Sunday February 02, 2025 @06:55PM (#65137281) Homepage Journal

    Imagine having a multi-billion-dollar valuation and then an open source product does 95% of what the customers want for free.

    It must be all hands on deck to keep the investors from going full Fukushima.

    Not an easy boat to row.

    • Re: (Score:2)

      by yog ( 19073 ) *

      Sort of like Linux versus Windows, back in the 1990s?

      I suspect it won't be long before OpenAI, Gemini, Meta, and Grok all have lightweight open sourced engines for hobbyists to play with. Grok's is already open source, actually.

      Also, when you say "does 95% of what customers want for free", do you mean DeepSeek's downloadable gguf files will somehow remain up to date if you install them on a local system?

      • Re: (Score:2)

        by edis ( 266347 )

        ...will somehow remain up to date if you install them on a local system?

        With the network available you don't care if the intelligence originates on your premises. You do not today, right?

      • Re: (Score:3)

        by SeaFox ( 739806 )

        and then an open source product does 95% of what the customers want for free.

        Sort of like Linux versus Windows, back in the 1990s?

        Um, what? Back in the '90s people wanted a computer that would run all the applications they heard about on TV and read about in magazines. They wanted a computer that "just worked" with a minimum of fiddling. Linux did neither of those back then.

        Today, Linux is a much more attractive choice because so much of what people do on a computer is through a web browser, and those are universal. We have things like Wine and other compatibility tools that allow some of those Windows-only productivity apps and games

    • Imagine having a multi-billion-dollar valuation and then an open source product does 95% of what the customers want for free.

      I guess it's news to many, there already were a bunch of free alternatives, Gemini, Grok, Meta, etc. on top of GPT being free to use. OpenAI isn't like Tesla pretending they've crossed the finish line bullshitting about their capabilities, it is what it is. AI right now isn't good enough, so where does 95% of not there yet get you?

      It's like making a 95% clone of a not-yet-cure for cancer that you could already get for free. 95% of a glider that still isn't an airplane. Fun to cheer the underdog I guess, but

    • I've been asking it unsuccessfully do some non-trivial task for a while now, but TODAY it did get it right!

      Q: I have a cylinder, diameter 24mm, 45mm tall, made of unknown material. The weight is about 211.5 grams. What could it be made of?

      ... a full minute later:

      A: To determine the possible material of a cylinder with the given dimensions and weight, we can start by calculating the volume of the cylinder and then use the density to find possible materials.
      Step 1: Calculate the Volume of the Cylinder
      Step 2: Calculate the Density
      Step 3: Identify Possible Materials
      Conclusion

      The calculated density of approximately 10.41 g/cm suggests that the cylinder could potentially be made of Lead or a material with a similar density. If the density were significantly higher or lower, we might consider other options based on specific applications, but from the values given, Lead seems to be the most likely candidate.

      If you have any additional specifications or context, let me know, and I can provide more detailed information!

      Man, that's Nobel Prize level stuff...

  • Poll: Should Slashdot ban 'twitter' links? (Score:5, Insightful)

    by Vegan Cyclist ( 1650427 ) on Sunday February 02, 2025 @07:27PM (#65137357) Homepage

    As an example, a huge number of subreddits have totally banned posting links to the Nazi-tolerant platform formerly known as twitter. It's run by 'the world's wealthiest man' who's a man-child that has achieved way too much power, and despite having the ability to do a tremendous amount of good, has instead decided to be a strong contributor to the division and degradation of society. Plus he cheats at video games.

    Kind of surprised to see a link in this post, but would be a good poll topic, and perhaps Slashdot can get on board with this.

    • Can Xwitter posts even been seen without an account?
      Last time I bothered looking I had to have an account, and then later it said my browser (Pale Moon) wasn't supported.

      Not that I feel like going back there after I removed all tweets and left a link to Mastodon on that infamous october 28th. Even before that time there were only like two out of several hundred people I "knew" to have an intelligent conversation with.

      • Can Xwitter posts even been seen without an account?

        ANY use of Xitter enables them to retain their position as an irreplaceable platform for breaking news / updates / whatever.

        It doesn't matter if I can view it logged out, ads disabled, proxied so they can't track me. Jumping through those hoops would just demonstrate that I'm still dependent.

        At this point, anyone who won't at least publish in parallel on another platform is not someone I need to follow.

        • Troll, really? I'll own that it's a somewhat spicy opinion, but it's sincere. I'm not just trying to start shit.

    • Re: (Score:2)

      by MpVpRb ( 1423381 )

      It's possible to get value from twit-x and avoid the crap.
      Follow trusted people. Only read their posts.
      Filter or ignore the crap. It's actually easy

    • Re: (Score:2)

      by Kokuyo ( 549451 )

      Well, your phrasing wants me to see more twitter posts now...

      Just kidding, I hate twitter. And since your post reminded me of it, I'd just like to see less of yours :).

    • As an example, a huge number of subreddits have totally banned posting links to the Nazi-tolerant platform formerly known as twitter.

      Of course, you started your thought in the subject line which makes quoting you seem weird... but this is expected from the type of person that would think to ban something.

      Only fascists and anti-fascists (which are really just fascists that are losing) ban things for other people.

      A person needs to log in to see anything twitter(sp) related. That means people like me will never see what is being pointed at; however, if it is the only source of information available, banning the links are effectively banning

      • > Only fascists and anti-fascists (which are really just fascists that are losing) ban things for other people.

        Are you seven years old? Because a 7yr old would be so ignorant as to make such an absolutist statement.

        So we shouldn't ban things, huh?

        So we shouldn't ban leaded gasoline?

        Asbestos should be free to use?

        Any pesticide, or food additive should be legal? And why not allow inclusion of these without mention - ingredient labels should be optional, or inaccurate, right?

        And child pornography? Who are t

        • Are you seven years old? Because a 7yr old would be so ignorant as to make such an absolutist statement.

          These are short-form, one sided comments. What do you expect?

          So we shouldn't ban things, huh?

          The context was information, not physical items, but again, this is short form, so explanatory words are usually wasted. Use your brain and keep the discussion within the context.

          So we shouldn't ban leaded gasoline?

          Asbestos should be free to use?

          Any pesticide, or food additive should be legal? And why not allow inclusion of these without mention - ingredient labels should be optional, or inaccurate, right?

          None of these are contextual, so, whatever. We could have an extended conversation about these out of context things, but that is beyond the scope currently.

          And child pornography? Who are these fascists behind banning these things anyway?

          Finally, something in context, unfortunately, it is a highly charged subject... but ok. I bet you would be surprise

  • We already know the answer (Score:5, Funny)

    by algaeman ( 600564 ) on Sunday February 02, 2025 @08:12PM (#65137441)
    We just need to figure out what the right question is. So long and thanks for all the fish.
  • I’m not convinced of the importance or difficulty of such breakthroughs when they seem to take a couple of days to a few weeks to complete, and only right after the stock price tanks.

    • Re: (Score:2)

      by edis ( 266347 )

      Depends on the cause of the tanking, too. If that's only due to the smart abuse, some manipulation trough misinformation and multiplication by the sheer number of the subscribers - how persistent it is?

      Actually, could it be, AI was requested to suggest a way of such tanking?

  • Just an idea for the lucrative branding.

  • Deep Seek or Deep Shit (Score:5, Insightful)

    by Required Snark ( 1702878 ) on Sunday February 02, 2025 @08:37PM (#65137475)
    The chaos caused by the Deep Seek unveiling is a stark reminder that AI/LLM is an out of control hype driven bubble. The valuations and the expectations (i.e. bullshit) are not objectively supported by real world results. It's about as realistic as Harry Potter's magic wand.

    If you have the depth of talent and resources of Google then you can win the Nobel Prize if you choose the right problem. Protein folding is an extremely bounded domain and the training data is of the highest quality, plus it reflect how chemistry works in the real world. It's as objective as it gets and results can be absolutely verified.

    When training data is scraped from the internet and the goal is make a billion dollars it's a cargo cult science [wikipedia.org].

    He drew an analogy to the cargo cult phenomenon in the South Pacific Ocean in which, as he understood it, islanders built a mock airstrip to cause airplanes loaded with imported goods to land. Similarly, he argued, adopting the appearances of scientific investigation without a self-critical attitude will fail to produce reliable results.

    Data scraped from the internet is about as clean as picking up the trash in an alleyway where drunks and drug addict pass their time. It's filled with garbage, excrement, used condoms, and hypodermic needles. Will you occasionally find a diamond ring in the trash? Yes, but that's neither science or a business model.

    The claims of "FREE MONEY" peddled by the AI hucksters are about as accurate as the six legged dog that pops out of a typical AI created image.

    • But crypto!

    • It is even confusing to me because iirc the differentiating factor between the original chatgpt and the other LLMs was exactly that it had been finetuned on clean data. As in, instead of scraping code snippets from the web, openAI hired coders to produce vetted code, etc.

      And that seemed to work, at least once! People were genuinely impressed by the change in quality. So if that is really the goal, why go back to scraping mountains of shit instead? There's something else at play, probably something stupid.

  • so.... (Score:3)

    by thanjee ( 263266 ) on Sunday February 02, 2025 @08:42PM (#65137481) Journal

    ...other than the bots, and Elon's alts, did anyone actually see this announcement on twitter?

  • ewwww (Score:3)

    by GrahamJ ( 241784 ) on Sunday February 02, 2025 @09:15PM (#65137527)

    They use xtwitter? That's terrible.

    Ban that site, block it at the firewall. The shitstain at the top deserves no attention.

  • Watching the tech bros of the world squirm as they are crushed by open source is a rare and delicate pleasure. The Altmans, Bezoses, and Musks of the world would strip us of all autonomy, self-determination, and common dignity—solely in pursuit of becoming the first trillionaire. Fuck them.

  • Because they accept non pre-paid credit cards as a means of payment exclusively.

  • I joked with my doctor prior to the procedure that he'd be doing deep exploration and learning.

  • PR. Look over here. (Score:3)

    by Big Hairy Gorilla ( 9839972 ) on Monday February 03, 2025 @12:04PM (#65138615)
    Now you can 't look over there .

    That is the attention economy.

    Don't downvote me, bro.

Slashdot Top Deals

The economy depends about as much on economists as the weather does on weather forecasters. -- Jean-Paul Kauffmann

Close