Next in Tech - Mar. 25, 2023

March 25, 2023

From the SEC warning to Coinbase and the arrest of Do Kwon, to the TikTok CEO's congressional hearing and OpenAI's ChatGPT plugin store, here's what's Next in Tech:

SEC Warns Coinbase

Coinbase has received a Wells notice from the SEC regarding an “unspecified portion of its listed digital assets, staking service, Coinbase Earn, Coinbase Prime, and Coinbase Wallet, after a cursory investigation”. This latest development is yet another blow in a year that has seen persistent negative sentiments towards crypto, further demonstrating the need for greater regulatory clarity and guidance.

A Wells notice is a way for the SEC to inform a company that they are recommending enforcement action for possible violations of securities laws. While the notice suggests that the SEC is contemplating enforcement action against Coinbase for its digital assets and services, it does not automatically signify that charges will be filed.

As outlined in a response from Chief Legal Officer Paul Grewal, Coinbase has expressed disappointment with the SEC's lack of clarity on which assets on its platforms the SEC believes may be securities. Coinbase had provided multiple proposals to the SEC about registration, all of which were ultimately refused, and the company claims that the SEC has not been fair or reasonable in its engagement on digital assets. The company expressed concern that the SEC's enforcement-only approach and lack of regulatory guidance for crypto companies will drive innovation, jobs, and the entire industry overseas.

Coinbase's Wells notice arrives during the same week as the president's annual economic report to Congress, in which negative sentiments were expressed towards crypto. As the report states, "crypto assets to date do not appear to offer investments with any fundamental value, nor do they act as an effective alternative to fiat money, improve financial inclusion, or make payments more efficient; instead, their innovation has been mostly about creating artificial scarcity in order to support crypto assets’ prices—and many of them have no fundamental value."

Crypto Charges: Do Kwon and Influencers in Trouble

Interpol has confirmed the arrest of Do Kwon, founder of the platform behind TerraUSD and Luna, in Montenegro. The South Korean national is wanted in the US and South Korea on fraud and other charges relating to the collapse of his crypto company, which wiped around $40bn from the crypto market and led to panic in May 2022.

Incidentally, this arrest only comes a day after the SEC charged crypto entrepreneur Justin Sun and his companies for securities fraud and other violations, including illegally paying eight celebrities to tout crypto asset securities Tronix and BitTorrent without disclosing their compensation. The celebrities charged include Lindsay Lohan, Jake Paul, Soulja Boy, Austin Mahone, Kendra Lust, Lil Yachty, Ne-Yo, and Akon.

TikTok CEO in Congress

TikTok CEO Shou Chew testified before Congress for over five hours on Thursday, facing questions about the privacy of US users and potential collusion between the app and Chinese officials. Members of the House Energy and Commerce Committee grilled Chew about TikTok's links to China, harmful content promotion, and the immense amount of data collected from users. While some lawmakers criticized the app, others focused on the need for industry-wide laws to protect children's online safety.

During the hearing, Congress also made the case that TikTok poses a national security threat due to the potential for the Chinese government to obtain the data of TikTok's 150 million US users or influence its recommendation algorithms to push propaganda or disinformation. Although there were concerns raised about TikTok posing a national security threat and potential data privacy risks, there was little concrete evidence presented to support these allegations. Instead, the hearing's main focus was advocating for regulations to protect American data privacy and prevent online harms, with TikTok being a primary subject of concern.

ChatGPT Plugins

It seems that every week OpenAI launches something that breaks the Internet. A few days ago, they unveiled a plugin store for ChatGPT that allows for web browsing, video functions, and connections to a range of popular apps, such as Kayak, Spotify, Expedia, OpenTable, Zapier, and many others. This breakthrough marks a major step forward in chat-based interfaces, providing users with a powerful tool to interact with their favorite apps using natural language. It's like having a command line that can now control our apps, giving us greater control over our online experiences.

OpenAI built a few starter plugins, such as web browsing and code interpreting. They also built a plugin called Retrieval, which can connect to personal or organization-wide datasets. This could be helpful for searching over documents, emails, or even notes. They have also released a waitlist and documentation for the plugin ecosystem, signaling how heavily they might want to lean into these features.

In the concept mockup above titled WalletGPT, a user can ask ChatGPT to find a song via a Shazam plugin, check if the song is collectable as an edition, and can purchase it with a hypothetical Rainbow Wallet plugin. Will chat be the interface of the future that interacts with every app we do? Whether the solution is all chat or not, there’s no doubt that AI advancements will dramatically change what it means to use the Internet in the next 5-10 years.

Following and thinking about the release of ChatGPT plugins has left me reflecting on two points:

  1. As someone who has seen and tried to build systems that connect data from different apps, I question whether the semantic lessons from the past are now outdated. Can a language model like a LLM act upon context-based data more efficiently, or are they simply new ways to access existing gateways to data? I wonder whether LLMs can improve current data communication methods independently of the LLM used to retrieve the data.
  2. Is chat going to win as an interface? Although chat has been revolutionary so far, I am uncertain whether its success is solely due to its conversational mode or rather because of AI's capabilities. If the latter is true, then I am intrigued to see how LLMs can transform new kinds of interfaces.

Interesting finds I’ve stumbled upon recently: