OpenAI’s DevDay And Fake Bard Ads — AI Brings Potential And Drawbacks

This is the published version of Forbes’ CIO newsletter, which offers the latest news for chief innovation officers and other technology-focused leaders. Click here to get it delivered to your inbox every Thursday.

In the year since OpenAI introduced ChatGPT to the world, AI has exploded. It’s risen from relative obscurity to become one of the top priorities for business spending throughout the world. And OpenAI, which received a major investment from Microsoft, is becoming one of the largest players in the space. News organizations have compared the company, which launched in 2015 as a nonprofit to pursue AI as a benefit for all humanity, to Big Tech.

At OpenAI’s first in-person DevDay conference last week, the company launched several new initiatives. The new GPT-4 Turbo has the ability to process up to 300 pages of text, meaning the application can handle more complex queries, but it can also bring more depth and context to responses. And the cost is just pennies. OpenAI also introduced a feature to allow users to personalize the ChatGPT bot, using knowledge for specific industries. These customized GPT bots could eventually be available for public download.

However, the new attention and features—plus intermittent outages from an old-school-style hacker attack—has led OpenAI to pause new signups for its paid ChatGPT Plus service. CEO Sam Altman tweeted the new interest “has exceeded our capacity and we want to make sure everyone has a great experience.”

It’s a reminder that AI technology is making huge strides, but it’s still a developing technology. There are still bumps to be figured out and problems to solve before it becomes the behemoth many are predicting.

LEGAL ISSUES

Earlier this week, Google sued three unnamed individuals in Vietnam, accusing them of making fake ads for its Bard AI chatbot. People who clicked the link downloaded malware that stole their social media login credentials instead of the chatbot. The lawsuit, filed in a California federal court, states that the company does not currently know the names and “capacities” of the defendants and is suing them under fictitious names. Google is seeking a court order to stop scammers from setting up fake profiles and wants to be able to disable fake pages registered in the U.S. If the case is successful, it could set a new precedent for U.S. tech companies to go after scammers around the world.

Omegle, a website that randomly paired strangers for online video chats, shut down last week, days after it reportedly settled a civil claim for damages in the online sexual abuse of a pre-teen girl. The site had been in operation for 14 years, and at one time had millions of daily users. But, as law enforcement agencies around the world found, it ended up pairing many children with sexual predators. In a statement posted to the site, Omegle founder Leif K-Brooks said “Omegle punched above its weight in content moderation.” Which it might have done, but it is difficult to police behavior on such a sprawling online platform.

CYBERSECURITY

Two-thirds of IT leaders surveyed by Rubrik Zero Labs said they fear the amount of data they have is growing at a faster rate than they can secure it. And while that problem may become more solvable with better enterprise-level data management practices, as companies take on new types of AI systems, the amount of information they need to protect will continue to increase.

In addition to new types of data, some data generally thought to be secure might not really be. A study by the Irish Council for Civil Liberties found the browser data of key government and military officials is being accessed by foreign states. The data—which can include location information, personal medical records, financial data and other private information—is available through the real-time bidding provision of online advertising. This system allows anonymized data to be shared between data brokers and advertisers, and is a major driver behind targeted advertising. This kind of information-sharing happens billions of times each day. However, it’s rather detailed and can be sold in a way in which key individuals can be identified. Following the report, the Irish Council for Civil Liberties is calling for the protocols for data sharing to be changed so more personal information is not included.

ARTIFICIAL INTELLIGENCE

Databricks, the software company that stores different kinds of data—structured and unstructured—in “lakehouses” that can be analyzed for powerful insights, has added another feature to its offerings: generative AI. The company bought MosaicML over the summer for $1.3 billion, and is layering gen AI on top of its “lakehouse” concept. Users will be able to use plain English—instead of a coding language—to query the data, making the kind of information the system can provide more available to all company leaders, not just those who work on the tech side.

“I think of the lakehouse as a small platform. It’s the foundation, and this is the skyscraper on top of it,” Databricks CEO Ali Ghodsi told Forbes.

Microsoft is getting into the AI hardware game. The company announced two new chips—the Microsoft Azure Maia 100 AI Accelerator and the Microsoft Azure Cobalt 100 central processing unit, which it said are optimized for generative AI and cloud computing. The chips will be available next year, and will compete with specially developed chips from Nvidia, as well as some announced by Amazon and Google. This announcement, further establishing Microsoft as a key AI player, helped continue the company’s stock rally kicked off by its strong quarterly earnings late last month.

Since the early days of the internet, members of the intelligence community have said that online writing, photos and videos have worked to radicalize people and inspire terrorist attacks. Microsoft is partnering with Tech Against Terrorism, an independent nonprofit organization launched by the UN in 2016, to develop an AI-powered tool to detect potentially harmful content. The pilot phase of this partnership will establish a framework to assess the accuracy of this kind of an AI bot and ensure it’s correctly flagging problematic content. If it is successful, the two plan to make it available to smaller platforms and nonprofits.

BITS + BYTES

Humane Debuts The $699 AI Pin, A Pocket-Sized Digital Assistant Without A Screen

Who needs screens? With its new AI Pin—a wearable smart AI device with a camera that can be controlled through voice, hand signals and motions, Humane is hoping many people will feel the same way. The company was founded in 2019 by former Apple execs Imran Chaudhri and Bethany Bongiorno, and the AI Pin officially launched last week. It can perform smartphone-like functions, including taking pictures, giving recommendations, playing music, providing information and translation. The camera provides the device with contextual clues, and the device has a laser projector that can display text on a person’s hand.

“We started this journey because we thought it was time for a shift,” Chaudhri told reporters last week at Humane headquarters in San Francisco. “The last era has plateaued.”

Whether the AI Pin catches on remains to be seen. While smartphones have been ubiquitous for years, other wearable devices, including VR headsets, glasses and cameras have not been embraced by the public just yet. However, the point of the AI Pin might just be that technology is crossing into a new phase: One in which AI operates everywhere all the time. Also notable is that the AI Pin requires no apps. It just operates as a device. On his Reddit account last week, Bill Gates wrote that this kind of interface is the future—a user tells their system what they want to do and can do it.

FACTS + COMMENTS

Elon Musk is looking at spinning off his satellite internet company Starlink as soon as next year—and possibly making it publicly traded—Bloomberg reported. Assets of the company are being moved from SpaceX into a subsidiary, the news service said.

Nearly 5,000: Number of small satellites providing internet service through Starlink as of July

$150 billion: Estimated worth of SpaceX, including Starlink

‘False’: Musk’s response to the story on X, the social network he owns formerly known as Twitter