AI and LLM News from October, 2023

Tom Hanks and Gayle King Were Victims of AI-generated Fake Video Advertisements

The likeness of Tom Hanks was stolen and used for a dental plan ad, while Gayle King’s deepfake video has her telling social media users about a weight loss program. Tom Hanks has used AI in the production of his own films and said he saw this coming. Hanks said, “Anybody can now recreate themselves at any age they are by way of A.I. or deep-fake technology.”

AI-powered Deepfake Porn is Out of Control

While it once took hundreds of images of a person’s face to create a convincing deepfake, it is now possible with just one or two images. When deepfakes were first created in 2017, they required significant computer processing power and some programming knowledge. Now, there are user-friendly deepfake applications available for iPhones. In other cases, deepfake creators take commissions, charging as little as $30 dollars to create explicit videos of a customer’s favorite celebrity, ex, or teacher.

As a result, both celebrities and average citizens are finding their faces inserted into pornography without their consent, and these videos are being uploaded to porn sites for anyone to see. One of the most prominent deepfake porn websites is averaging about 14 million hits a month.

The proliferation of child sexual abuse images on the internet could become much worse if something is not done to put controls on artificial intelligence tools that generate deepfake photos. The Internet Watch Foundation urged governments and technology companies to act quickly before a flood of AI-generated images of child sexual abuse overwhelms law enforcement investigators.

TOP

Baidu Releases Generative AI ERNIE 4 Bot While the US Throttles AI-Training Chip Exports to China

Baidu received approval to release the latest version of its generative AI bot to the public, after nine months of being available only to corporate clients and select individuals. Like ChatGPT, ERNIE can generate answers to questions, but the Chinese government has published rules mandating that generative AI tools abide by the same censorship as social media. For example, it’s illegal for a chatbot to discuss the Chinese Communist Party’s (CCP) ongoing persecution of Uyghurs and other minorities in Xinjiang. And if you ask ERNIE what happened in China in 1989, instead of reporting that the Chinese army massacred thousands of pro-democracy protesters in Tiananmen Square, it will say it has no relevant information.

The US imposed restrictions on semiconductor exports to China in 2015, extended them in 2021 and twice in 2022, and this year added new restrictions to stop the flow of advanced chips to Chinese data centers that house supercomputers and infrastructure that supports AI development.

TOP

AI Chatbots Replacing Call Center Agents Worldwide

Starting around Y2K, western companies couldn’t offshore customer service jobs to India fast enough. But now, call centers are outsourcing to AI chatbots, which will profoundly effect economies – especially in countries like India and the Philippines. In India, a call center manager replaced his entire customer service department with a custom chatbot trained on the company’s call center data. He said in an interview:

“It was [a] no-brainer for me to replace the entire team with a bot, which is like 100 times smarter, who is instant, and who cost me like 100th of what I used to pay to the support team.”

TOP

Stack Overflow Lays Off 28% of Staff Due to ChatGPT’s Coding Prowess

Like many tech companies, Stack Overflow went on a hiring spree during the COVID-19 pandemic, doubling its headcount in 2022. But with the release of ChatGPT, it’s now laying off 28% of its workforce to cut costs while it develops its own coding chatbot called Overflow AI. As a forum for coders, Stack Overflow can’t compete with ChatGPT, which can generate and correct code, and provide optimization suggestions, and explanations of what each line of code is doing. Ironically, ChatGPT developed its coding prowess by scraping Stack Overflow’s forums – for free.

Billionaire-Funded Positions in Government Offices are Shaping the Future of AI Regulations

Open Philanthropy, which is financed primarily by billionaire Facebook co-founder and Asana CEO Dustin Moskovitz and his wife Cari Tuna, is funding the salaries of more than a dozen AI fellows in key congressional offices, in federal agencies, and at influential think tanks. AI fellows, working through the Horizon Institute for Public Service, which is a nonprofit created by Open Philanthropy last year, are present at the Department of Defense, the Department of Homeland Security, the State Department, the House Science Committee, and Senate Commerce Committee, Senators’ offices, and the RAND think tank. They are focusing on the long-term threats that future AI systems might pose to human survival, instead of immediate threats that AI is used to inject bias, spread misinformation, threaten copyright protections, and weaken personal privacy.

TOP

Bard Improvements

Bard can now respond with real-time info from Google Maps, Flights, Hotels, and YouTube, and can find and bring in information from Gmail, like the time of an event or a note from a boss.

Bard can generate summaries of content from documents and PDFs, including key points from content across Google Docs and Google Drive.

Bard still hallucinates terribly, so now you can doublecheck its statements by using the Google Search button to highlight search results content that convfirm or contradict its statements. These highlighted results are clickable to learn more from the sourced website.

OpenAI Integration of ChatGPT and DALL-E is Inconsistent With Limited Capabilities

With a ChatGPT Plus account, you can upload and analyze an image, but not create a new one, while the Enterprise version allows a user to create an image from within ChatGPT, but not upload one. Meanwhile, with DALL-E you can upload and edit an image, but not analyze it or create a similar one. Hopefully, image upload, analysis, and creation will come together soon. Both Plus and Enterprise versions now feature web browsing with Bing, support for plugins, and advanced data analysis.

In addition to seeing and analyzing images, ChatGPT Plus can now read and pull important points out of PDFs, edging it towards the functionality of the Enterprise version. And with the Advanced Data Analysis feature, both Plus subscribers and Enterprise users can work with spreadsheets and ask ChatGPT to analyze them and pull our pertinent information.

ChatGPT Downloads and Revenue Continue to Grow Along With AI Image Apps

With 3.9 million downloads in its first month, the ChatGPT mobile app has grown to 23 million downloads as of September 2023, with a monthly revenue of $2.39 million as of October 24 – and that’s just the mobile app. But mobile apps competing with ChatGPT such as Chat & Ask AI, ChatOn — AI Chat Bot Assistant, AI Chatbot — Nova, and AI Chatbot: AI Chat Smith all have millions of monthly downloads and generate more than $1 million in monthly revenue.

ChatGPT still leads the pack, but some photo editing AI apps have gone viral. Remini topped 16.17 million downloads last month after its AI headshots feature blew up on TikTok. And the AI-powered PicsArt AI Photo Editor had nearly 16.05 million installs in September.

TOP

Character.AI Introduces Group Chats Where People and Multiple AIs Can Talk to Each Other

The chatbot platform developed by ex-Google AI researchers offers customizable AI companions with distinct personalities and tools to make your own is now offering a group chat experience where users and their friends can chat with multiple AI characters at once. For example, users can chat with Albert Einstein, Marie Curie, Nikola Tesla, and Stephen Hawking, or form a group of friends to discuss travel with an AI travel agent augmenting the conversation. Group chat is available on iOS and Android with a $9.95 monthly subscription, with web access rolling out soon. Having a user talk to different bots, or a team converse with a chatbot, is supposed to aid creativity and productivity, while it’s hoped that bots chatting with each other may lead to even more advances in AI and solutions to other problems.

Group sessions with AI chatbots are also available on Snapchat, and across Meta’s apps including WhatsApp, Messenger, and Instagram.

TOP

Google’s AI-Enhanced Search Now Generates Images and Writes Drafts

Google search generative experience (SGE) is looking more like Bing Chat with DALL-E built in. Google SGE introduced the ability to generate images using prompts directly in SGE similar to rival Bing’s support of OpenAI’s DALLE-E 3. In addition, SGE will now allow users to write drafts within SGE and customize the output to be longer or shorter or change the tone of the writing to be more serious or casual.

President Biden Issues Executive Order on AI

In what has been called a broad initiative without enforcement mechanisms, President Biden issued an executive order aimed at safeguarding against threats posed by artificial intelligence, and ensuring that bad actors do not use the technology to develop devastating weapons or mount supercharged cyberattacks. The executive order exerted oversight over safety tests that companies use to evaluate conversation bots such as ChatGPT and introduced industry standards like watermarks for identifying AI-fueled products, among other regulations. The batch of orders included the following:

• Require that developers of the most powerful AI systems share their safety test results and other critical information with the U.S. government.
• Develop standards, tools, and tests to help ensure that AI systems are safe, secure, and trustworthy.
• Protect against the risks of using AI to engineer dangerous biological materials.
• Protect Americans from AI-enabled fraud and deception by establishing standards and best practices for detecting AI-generated content and authenticating official content.
• Establish an advanced cybersecurity program to develop AI tools to find and fix vulnerabilities in critical software.
• Order the development of a National Security Memorandum that directs further actions on AI and security.

The executive order also addressed privacy protection, stating, “Without safeguards, AI can put Americans’ privacy further at risk. AI not only makes it easier to extract, identify, and exploit personal data, but it also heightens incentives to do so because companies use data to train AI systems.” the President calls on Congress to pass bipartisan data privacy legislation to protect all Americans, especially kids, and directs the following actions:

• Protect Americans’ privacy by prioritizing federal support for accelerating the development and use of privacy-preserving techniques.
• Strengthen privacy-preserving research and technologies.
• Evaluate how agencies collect and use commercially available information.
• Develop guidelines for federal agencies to evaluate the effectiveness of privacy-preserving techniques.

The executive order also addressed ambitions to limit algorythmic discrimination, aims to stop unsafe AI practices in healthcare, support workers, and issue guidelines for government use of AI. While the order has more teeth than the voluntary commitments Biden has secured from some of the biggest AI companies, many of its provisions don’t and can’t have the force of law behind them, and their effectiveness will largely depend on how the federal agencies named within the order carry them out. They may also depend on if those regulations are challenged in court.

TOP