Within minutes, the bot had him furious. Wadhwa told The Tech 202 by phone Monday that he first asked who he was, then what he was famous for, and finally what he was “infamous for.” In response to the final question, Bott described his 2014 blog post accusing Wadhwa of plagiarism, with a byline, date, and official citation. As you might have guessed at this point, there are no blog posts.
Nowadays, “hallucination” is a familiar problem for AI chatbots.my colleague Pranche Verma And I reported on a similar episode last year where OpenAI's ChatGPT fabricated a sexual harassment scandal against a law professor. Jonathan Turley. Since ChatGPT was released in late 2022, there have been countless reports that ChatGPT and other “generative AI” tools from industry heavyweights like Microsoft and Google have spewed misinformation, displayed bias, and taken inappropriate actions. Reports have been received.
But the risk of chatbots failing doesn't seem to be slowing tech giants' efforts to integrate chatbots into the world's largest online platforms.
Meta recently began rolling out its Meta AI chatbot more broadly in India and select parts of Africa, including on WhatsApp, Instagram, and Messenger, bringing the tool to markets serving hundreds of millions of people. I am. And on Monday, new york times The company reported that it is approaching American Instagram influencers about a project called “Creator AI,” in which AI avatars trained to use their voices will interact with fans on their behalf.
For a company that has always prided itself on its AI research but has lagged behind rivals Microsoft, Google and OpenAI in commercializing the latest language and image models, the move marks the start of AI technology for people in their daily interactions. represents an ambitious leap forward in integrating the When Meta announced Meta AI in September, the company said it envisioned people asking the chatbot questions, using the chatbot to create images, or just chatting.
Given what we know so far about chatbots, this is hubris.
AI chatbots amaze with their ability to converse on a wide range of topics and generate images of everything people dream of (albeit with some limitations). But what the industry calls “hallucinations” — reactions that have no basis in reality — are starting to look more like a feature than a bug. More recently, companies including Microsoft, Google, OpenAI, and now Meta have tried to solve the problem by connecting AI models to the web, allowing them to search for information and cite sources. It was a mixed bag. On the other hand, attempts to train chatbots to eliminate aggressive bias detected from training data often fail or go too far.
Meta is not alone in integrating chatbots despite the risks. Last year, rival Snapchat launched My AI and My Colleagues, billed as friendly conversation partners. Jeffrey A. Fowler It quickly became clear that I was giving inappropriate advice to young people. Google and Microsoft are increasingly incorporating AI language tools into their search engines and productivity tools, but Microsoft is being used to generate pro-Nazi memes and Google is overcorrecting to avoid racial stereotypes. Even though there was a tendency to do so. Elon MuskX recently leveraged Grok AI tools to create trending headlines, some of which were quickly proven to be false. Even Amazon released an AI chatbot for shopping. My colleague took advantage of it. Shira Obide It was reviewed as “stupid as a stone”. (Amazon founder jeff bezos Owner of the Washington Post. )
At this point, it seems like the meta AI is fabricating the controversy regarding Wadhwa, which is not surprising.
As with other tools so far, Meta's response is that these tools are a work in progress.
“Meta AI and our suite of generative AI capabilities help increase creativity and productivity for people using our apps,” a spokesperson said. Kevin McAllister said in an emailed statement. “As we said when we announced these new features in September, this is new technology and it won’t always be perfect. This is true of all generative AI systems. We continually release updates and improvements to our model and continue to work to make it even better.”
Wadhwa said he disagreed with such a defense. “WhatsApp is already being used in India to influence elections and to incite hatred and misinformation,” he said. “Now imagine an AI chatbot that generates very trustworthy-sounding references that people trust. …This is dangerous.”
Matt Navarra, a UK-based social media consultant and industry analyst, said that despite these downsides, tech giants see AI chatbots as strategically important.
“What is likely going to happen is that most platforms, like TikTok and That means you'll be able to do a search.'' Navarra said. While people may find this feature “gimmicky” for now, tech companies can collect users' personal data, keep it on their platforms, and use those interactions to collect data for future applications. provides another way to further train the AI system. So that someday it will become “something more meaningful.”
In the meantime, he said, expect “many more headlines” about bias and mishandling, as these issues are unlikely to be resolved anytime soon.
Report finds social media tools are inadequate to track political ads
Tools developed by social media companies to help people track political ads on their platforms are all “plagued by missing data, bugs, poor functionality, and unacceptable deficiencies,” a new report finds. did.my colleague gerrit de vink We report on findings from Mozilla, the maker of the Firefox web browser, and CheckFirst, the Finnish internet transparency organization.
Over the past few years, companies like Meta, Snapchat, TikTok, and Google have been building “ad repositories” that allow researchers and journalists to see who is paying for political ads on these platforms. Ta. Regulations set to take effect in the European Union later this year will also require companies to be transparent about their political advertising funding.
However, an investigation by Mozilla and FirstCheck found that Snapchat, X (formerly Twitter), and Microsoft Bing's ad transparency tools are all effectively unusable. Google's tools had “minimum data and functionality,” while Apple's app store, LinkedIn, Meta, and TikTok tools had “significant gaps in data and functionality.” The report said none of the platforms were “ready for action.”
The tool has become an important data source for activists and researchers as billions of people head to vote this year around the world, including in the United States and India.
“Our research shows that most of the world's largest platforms do not offer functionally useful ad repositories,” he said. Claire Parshan, Head of EU Advocacy at Mozilla. “The current suite of tools does exist, but in some cases that is all that can be said about them.”
Battle lines being drawn as US states confront big tech companies over online child safety legislation (The Guardian)
X changes policy and tells Brazilian court it will abide by the ruling (Reuters)
Judge halts Texas investigation into Media Matters' X coverage (Ars Technica)
Adobe considers OpenAI partnership as it adds AI video tools (Reuters)
Microsoft to invest $1.5 billion in Abu Dhabi AI group G42 (Financial Times)
SpaceX may finally face competition. It may already be too late. (Written by Christian Davenport)
Meta will temporarily shut down Threads in Türkiye on April 29th. (The Verge)
Broadcom faces questions from EU over VMware license changes (Reuters)
Scammers are targeting teenage boys on social media, leading some to commit suicide. (Bloomberg)
They criticized Israel. This Twitter account changed their lives forever. (Written by Pranche Verma)
- The Stanford Human-Centered AI Institute (HAI) will hold an event, “AI+ Policy Symposium: Global Stocktaking,” at noon today.
- The Data Innovation Center will be hosting the event 'How can the UK accelerate the adoption of AI in the public sector?' from 3pm today.
- Semaphore will host its World Economic Summit event on Wednesday and Thursday at Gallup's Great Hall and Mellon Auditorium.
that'That's all for today — thanks for joining us! Tell others to subscribe of technology 202 here.Please contact Cristiano (email or Social media) and will (email or Social media) Please give us any tips, feedback, or say hello!
