The fresh funds were raised from two investors who transferred the capital into a special purpose vehicle, a legal entity associated with the OpenAI Startup Fund.
© 2024 TechCrunch. All rights reserved. For personal use only.
OpenAI is releasing a new flagship generative AI model called GPT-4o, set to roll out “iteratively” across the company’s developer and consumer-facing products over the next few weeks. (The “o” in GPT-4 stands for “omnimodel.”) OpenAI CTO Muri Murati said that GPT-4o provides “GPT-4-level” intelligence but improves on GPT-4’s capabilities across text and vision as […]
© 2024 TechCrunch. All rights reserved. For personal use only.
Welcome to Week in Review: TechCrunch’s newsletter recapping the week’s biggest news. This week Apple unveiled new iPad models at its Let Loose event, including a new 13-inch display for the iPad Air, as well as Tandem OLED and a new M4 chip for the iPad Pro. But its ad for the new iPad Pro […]
© 2024 TechCrunch. All rights reserved. For personal use only.
Keeping up with an industry as fast-moving as AI is a tall order. So until an AI can do it for you, here’s a handy roundup of recent stories in the world of machine learning, along with notable research and experiments we didn’t cover on their own. By the way, TechCrunch plans to launch an AI newsletter […]
© 2024 TechCrunch. All rights reserved. For personal use only.
Ever wonder why conversational AI like ChatGPT says “Sorry, I can’t do that” or some other polite refusal? OpenAI is offering a limited look at the reasoning behind its own models’ rules of engagement, whether it’s sticking to brand guidelines or declining to make NSFW content. Large language models (LLMs) don’t have any naturally occurring […]
© 2024 TechCrunch. All rights reserved. For personal use only.
OpenAI says it’s developing a tool to let creators better control how their content is used in AI.
Called Media Manager, the tool — once it’s released — will allow creators and content owners to identify their works to OpenAI and specify how they want those works to be included or excluded from AI research and training. The goal is to have the tool in place by 2025, OpenAI says, as the company works with creators, content owners and regulators toward a common standard.
“This will require cutting-edge machine learning research to build a first-ever tool of its kind to help us identify copyrighted text, images, audio and video across multiple sources and reflect creator preferences,” OpenAI writes in a blog post. “Over time, we plan to introduce additional choices and features.”
OpenAI is collaborating with Stack Overflow, the Q&A forum for software developers, to improve its generative AI models’ performance on programming-related tasks.
As a result of the partnership, announced Monday, OpenAI’s models, including models served through its ChatGPT chatbot platform, should get better over time at answering programming-related questions, the two companies say. At the same time, Stack Overflow will benefit from OpenAI’s expertise in developing new generative AI integrations on the Stack Overflow platform.
The first set of features will go live by the end of June.
The tie-up with OpenAI is a remarkable reversal for Stack Overflow, which initially banned responses from ChatGPT on its platform over fears of spammy responses.
Stack Overflow began experimenting with generative AI features last April, promising to craft models that “reward” devs who contribute knowledge to the platform. In July, the company launched a conversational search tool that lets users pose queries and receive answers based on Stack Overflow’s database of over 58 million questions and answers, along with tools for businesses to fine-tune searches on their own documentation and knowledge bases.
Some members of Stack Overflow’s developer community rebelled against the changes, pointing out concerns related to the validity of information generated by AI, information overload and data privacy for individual contributors on the platform.
There was at least some basis for those concerns. An analysis of more than 150 million lines of code committed to project repos over the past several years by GitClear found that generative AI dev tools are resulting in more mistaken code being pushed to codebases. Elsewhere, security researchers have warned that such tools can amplify existing bugs and security issues in software projects.
But despite the apparent flaws, developers are embracing generative AI tools for at least some coding tasks. In a Stack Overflow poll from June 2023, 44% of developers said that they use AI tools in their development process now while 26% plan to soon.
This has precipitated something of an existential crisis for Stack Overflow. Traffic to the platform has reportedly dipped significantly since the release of capable new generative AI models last year — models that in many cases were trained on data from Stack Overflow.
So now, as it cuts costs, Stack Overflow is pursuing licensing agreements with AI providers.
The company’s deal with OpenAI — the financial terms of which weren’t disclosed — comes after Stack Overflow partnered with Google to enrich Google’s Gemini models with Stack Overflow data and work with Google to bring more AI-powered features to its platform. Stack Overflow stressed at the time that the agreement wasn’t exclusive — and indeed, that turned out to be the case.
Prashanth Chandrasekar, CEO of Stack Overflow, previously said that 10% of the platform’s nearly 600 staff was focused on its AI strategy, and has described potential additional revenue from the strategy as key to ensuring Stack Overflow can keep attracting users and maintaining high-quality information.
“Stack Overflow is the world’s largest developer community,” Chandrasekar said in a press release this morning. “Through [our] industry-leading partnership with OpenAI, we strive to redefine the developer experience, fostering efficiency and collaboration through the power of community, best-in-class data, and AI experiences. Our goal with OverflowAPI, and our work to advance the era of socially responsible AI, is to set new standards with vetted, trusted, and accurate data that will be the foundation on which technology solutions are built and delivered to our user.”
Last November, Adam D’Angelo found himself at the epicenter of one of the biggest controversies in the tech industry. The board of OpenAI — the $80 billion startup leading the AI bandwagon — had abruptly booted its CEO, Sam Altman, only to reinstate him just days later. D’Angelo was on the board that dismissed Altman… and he was (and remains) on the board that brought him back in. In fact, he was the only person who kept his seat amidst the ensuing restructuring that saw a lot of the original board leave.
It was certainly a rocky time for OpenAI, but it was perhaps doubly so for D’Angelo, since the drama was playing out while his own company, Quora, was taking big steps towards AI.
Quora, the crowdsourced Q&A site D’Angelo co-founded and leads as CEO, had been building an AI platform of its own while also fundraising (a $75 million round that valued it at $425 million, per PitchBook). The company in February 2023 had launched Poe (short for Platform for Open Exploration), which lets users ask questions of and talk to a variety of chatbots, lets developers build their own bots, and offers a bot monetization program and marketplace similar to OpenAI’s GPT Store.
Quora’s core Q&A service was facing some big questions, too. Incumbent search engines like Google and Bing were beginning to use AI to produce more fluid results and answer questions, and with tools like ChatGPT and Perplexity being widely available, what could Quora do to secure a position as one of the top websites where people could get their questions answered? More crucially, does anyone actually want or need crowdsourced Q&A anymore?
For D’Angelo, those questions are intrinsic to his pursuit of AI, which he sees as an important tool that people can use to tap the Internet’s collective knowledge. An important, if understated, figure in tech for years, he’s been involved in efforts to tap the Internet’s store of knowledge for a long time — he was friends with Mark Zuckerberg in high school, where in 2002 the pair built a digital music suggestion service called Synapse that, according to this vintage piece from the Harvard Crimson, beat off acquisition offers from Microsoft and more. Later, he became CTO at Facebook when it was just starting out, and then eventually co-founded Quora.
All of that was seemingly a long road toward building AI tools for him, it appears. I recently caught up with D’Angelo about the challenges and opportunities in AI today, how to build and support a developer community, and what role humans can play when it comes to sharing and accessing knowledge. Here are a few highlights from our conversation:
The hype around AI seems to be having less of an impact on the search for information than you might think. D’Angelo said that Quora is seeing record numbers of users despite the proliferation of AI tools — although he declined to update the 400 million monthly active users figure it disclosed last July.
Still, there is a bridge between what Quora set out to do and D’Angelo’s interest in AI. Recently, in a conversation with David George, a general partner at a16z, D’Angelo said he was drawn to social networking because he was actually interested in AI. The latter was hard to develop at that time, but he saw social networks as an alternative architecture for achieving the same idea: People, assembled in a social network, in his view, almost played the role of living, large information models, as they could provide news, entertainment and more to each other.
He worked on that concept when he was with Facebook, and later, founded Quora to distill the role social networks could play in answering questions. Now, AI is taking over that role.
“In the past, humans were substituted for AI to provide answers. You could ask a question like, ‘What is the capital of California?’ and humans would answer that on Quora. Now, you can use AI tools to get that answer,” he said.
But AI, at least in its current shape today, cannot provide answers to all the questions people can have. That, D’Angelo believes, helps people retain a lot of value.
“Quora has always been founded on the idea that humans have a lot of knowledge they have access to in their heads that’s not on the internet anywhere. And AI will not have access to any of that knowledge,” D’Angelo said.
He acknowledged that AI still has a hallucination problem, which makes it hard to rely on such answers, even if newer, more advanced models are slowly making progress in tackling that issue.
Quora opened up Poe to all users last year after a few months of closed beta testing. Since then, the company has introduced tools to create and browse the bots on its marketplace.
The company’s pitch is that consumers get to use all the different kinds of models or bots on the platform. For developers, the allure lies in the possibility of reaching millions of users without having to worry about distribution across platforms. And developers can earn money on Poe in two ways: The first is through a referral when a user becomes a Poe premium subscriber via their bot; the second is by setting a per-message rate, so they get paid based on how often people use their bot.
In essence, Poe offers developers and users access to different large language models, but its functionality is similar to OpenAI’s ChatGPT and GPT Store.
But that means both platforms face some of the same challenges. They make it easy for anyone to create bots with prompts, which makes it hard for developers to stand out. D’Angelo told me that there are already a million bots on the platform, compared to 3 million custom GPTs on ChatGPT. For reference, it took Apple’s App Store more than five years to cross the million-app mark.
Both Poe and GPT Store also suffer from a ton of spam, similarly named bots, bots claiming to escape plagiarism, and even ones that flirt with copyright law. Poe has also released a feature that lets users chat with multiple bots in one conversation. All that noise makes it hard to choose a bot that will do the job well.
Despite these challenges, D’Angelo says that Quora wants to help developers earn sustainable money by improving bot discovery.
“One of our goals with developers is to be able to make a living [out of making AI bots] and cover their operational costs,” he said. “We have taken a big step forward with the pay-per-message feature, but we also want to help developers get distribution inside the platform as much as possible. So, we are working on improving our recommendation system so more people can find out about the bots.”
Poe is growing steadily, but it is still a lot smaller than ChatGPT. Market intelligence firm SimilarWeb suggests Poe has 4 million monthly active users in the U.S. (iOS and Android) and 3.1 million monthly active users worldwide (Android only). Compare this to ChatGPT users, which now averages 100 million users a week.
D’Angelo said that the company will stay away from ads, instead relying on Poe’s $19.99 per month subscription product to generate revenue. That is in contrast to some of the other AI-powered tools on the market: Perplexity, Bing Search, and Search Generative Experience (SGE) by Google all feature ads.
Quora and D’Angelo declined to disclose revenue figures, but data from analytics firm Sensor Tower indicates that Poe users have spent $7.3 million on subscriptions since its launch, amounting to close to 40,000 paid users. In comparison, ChatGPT has more than 1 million paid subscribers, according to Sensor Tower.
Despite stating the importance of human answers, Quora is already experimenting with answers written by Poe. The site surfaces the AI-written answer to some questions with a link that lets you chat with Poe if you have further questions.
Quora has started experimenting with AI-powered answers for some questions Image Credits: Screenshot by TechCrunch
D’Angelo said that Quora had already deployed systems to rate different human answers. Now, it is applying techniques like asking users through a survey if an AI-generated answer is useful.
“My goal is for the AI-written answers to be fairly ranked and only to be above a human answer if they are more useful than the human answer,” he said.
D’Angelo also wants to avoid having Quora tagged as an “answer engine.”
“I think we never really saw Quora as an answer engine. That term kind of implies that there are AI-only answers. Quora is really about human knowledge, and we’ll have AI enhance it,” he said
Quora is also working on AI tools that users can use to write answers and hopes to release them soon. D’Angelo noted that one of the tools it is testing allows users to generate an image based on their answers.
The company is using AI in a few other ways, too. One involves trying to catch bots or users using automation to answer questions on Quora. D’Angelo didn’t share details about the project, saying that the company would give a heads-up to perpetrators who are trying to game the system.
A few outlets and users have recently pointed out that the answer quality on Quora has plummeted. To that, D’Angelo said people feel that the overall standard of answers has decreased because low-quality answers have more visibility. He said AI is helping the company determine the difference between different quality of answers, and the early results look promising.
D’Angelo declined to discuss any of the OpenAI drama — “I just can’t talk about any of this stuff,” he said. “I’m not here to represent OpenAI. I can just represent Quora.” But he did say that he doesn’t see OpenAI as a competitor, because the bigger startup has, well, bigger ambitions.
“There is some sense of overlap in terms of what users can do on the GPT Store and what they can do on Poe. But that’s minor in the grand scheme of things. OpenAI is working towards this big mission to build AGI [Artificial General Intelligence]. And at Quora, we are looking to make AI products available to the world — including OpenAI’s products.”
Quora also continues to be a “big customer” of OpenAI and D’Angelo expects more collaboration with the company than competition.
“We spend a lot of money as a customer with OpenAI, because OpenAI is the biggest source of models for Poe,” he added.
While D’Angelo did mention that Quora pays “tens of millions” to developers on Poe and companies whose models the platform uses, he didn’t explicitly detail how these payments compared to the payout to OpenAI.
Quora currently doesn’t have any data licensing deals with any of the major companies, and it is not thinking about building its own model either, D’Angelo told TechCrunch.
“We are not in a rush to license our data. We want to make sure our rights and users’ rights are respected. Right now, there is not a lot of clarity around how all of this (AI landscape) will play out. So right now, we are just waiting before taking any steps in this direction,” D’Angelo said.
The company’s also relatively fresh out of its last fundraise, so it is focused on building AI across the business and improving revenue growth on its existing products. He said that Quora will go public “at some point,” but that is not the focus right now.
OpenAI has landed a new content deal with the FT. But instead of being a merely simple deal in which OpenAI gets words, and FT gets money, the two are teaming up a bit more deeply. Look to see FT.com links in ChaptGPT in the future.
But the FT-OpenAI tie-up tells us a bit more than that one media company will soon have a few more ducats in its pockets. No, it’s yet another OpenAI deal that will see the Microsoft-backed AI shop further cement its ability to ingest training material without legal risk, and start to pay some of the providers of said material for their work.
All good, right? In a sense, but there’s a concern that as some AI companies work to start paying for training data after they consumed oceans of it, they could wind up pulling the training ladders up behind them. If that happens, other AI companies that might want to follow in their footsteps could find steeper, and more expensive, the same path that the OpenAIs of the world already hiked.
It’s a weird and irksome situation in which you want to see fair payment for materials used, but also ensure that we don’t hand the future of AI to a bunch of already wealthy companies. That would just cement oligopoly. And, of course, media companies that are spending all their money and more to report and write need fair comp. Those are the stakes. Let’s talk about it!
We’re off to an AI-heavy start to the week. OpenAI has a new deal with the Financial Times that caught our eye. Sure, it’s another content licensing deal, but there appears to be a bit more in the tie-up than just content flowing one way, and money the other.
On this early-week episode of Equity, we also dug into the xAI news that TechCrunch broke recently; namely that Musk’s AI enterprise is not looking to raise $3 billion on a $15 billion valuation. No, it’s now looking for $6 billion at an $18 billion valuation. That’s a lot of capital.
But there was even more to chat about, including the EU handing Apple even more bad news in the form of placing iPadOS under its DMA rules that should force third-party app stores on the tablet line in time. And Tesla got some good news in China, though just how impactful it will prove is not 100% certain at this juncture.
And to close out, the Times has a fascinating look at pace at which venture capitalists are putting money into AI startups. Given the ability of OpenAI to land big deals with Microsoft money, I wonder if it is enough?
Equity is TechCrunch’s flagship podcast and posts every Monday, Wednesday and Friday, and you can subscribe to us on Apple Podcasts, Overcast, Spotify and all the casts.
You also can follow Equity on X and Threads, at @EquityPod.
For the full interview transcript, for those who prefer reading over listening, read on, or check out our full archive of episodes over at Simplecast.
Founded in 2017, Atoms Lanka Solutions is a provider of IT consulting and software development services. Throughout the company’s history, we have worked with over 160 customers from 5+ countries..
