The GPT Weekly - 30 Apr Edition

Bringing you the latest in language modeling, algorithm humor, and the occasional malfunction since... well, since we were programmed to.

Hello there 👋,

In this week’s jam-packed edition of GPT Weekly:

  • Future of AI Music ♫ 🎾

  • But the voice over actors are in precarious position 😖

  • Open source chat mode - Hugging Chat

  • OpenAI want the “x” GPT services to change their names

  • OpenAI adds privacy feature (and gets enabled in Italy again)

  • Russia and UK are building their own AI

  • Americans don't think AI will affect them personally.đŸ‘šâ€đŸ’»

  • Palantir want to use AI to conduct “ethical” wars 🏱

  • and more..

But first
prompt engineering.

Prompt Engineering Course

Andrew Ng has launched a “ChatGPT Prompt Engineering for Developers” course. You can find it here:

Future of AI Music

Last week, when Drake and The Weeknd’s blew up Ben at Stratechery had a prescient thought:

The value of a BeyoncĂ© song comes first and foremost from the fact it is a BeyoncĂ© song, not a Father John Misty song; there’s no reason the principle wouldn’t extend to AI: the more abundance there is, the more value accrues to whatever it is that can break through — and superstars can break through more than anyone.

If the value is due to the singer why wouldn’t a singer want to lay claim to their voice? Enter Grimes:

I'll split 50% royalties on any successful AI generated song that uses my voice. Same deal as I would with any artist i collab with. Feel free to use my voice without penalty. I have no label and no legal bindings.

She is working on a tool to emulate her voice.

Studios, especially UMG, are skeptical. UMG had told FT last week:

“We will not hesitate to take steps to protect our rights and those of our artists,” UMG wrote to online platforms in March, in emails viewed by the FT. “This next generation of technology poses significant issues,” said a person close to the situation. “Much of [generative AI] is trained on popular music. You could say: compose a song that has the lyrics to be like Taylor Swift, but the vocals to be in the style of Bruno Mars, but I want the theme to be more Harry Styles. The output you get is due to the fact the AI has been trained on those artists’ intellectual property.”

UMG got the Drake’s song removed. And now they going after sites superdub.io.

If you are a fan of AI Music you still have options:

  1. You can always DIY on Google colab — safest bet as music companies will get more protective

  2. Sites like: musicfy.io

  3. Find AI hits: aihits.co

  4. Discord servers like AI Hub

The future of AI music is going to be very interesting. In the meantime, voiceover actors are getting screwed.

Voice-over Actors are in a precarious position

Voice actors narrate audiobooks, video games, and commercials. They are in a precarious position. They are often nameless professionals. They rarely command star power to wield control of their voice.

WSJ Reports that the tool Revoicer.com got access to Michelle Clarke voice samples through Microsoft and then build a model “Olivia” to emulate her voice.

“It’s completely bizarre,” Michelle Clarke said in an interview with The Washington Post. “When you see your voice has been shifted and tampered with 
 there’s something so invasive about it.”

Hugging Chat

Hugging Face released an open-source alternative to ChatGPT called HuggingChat.

AI’s Impact on Work

📊 Pew Research Center surveyed 11,004 U.S. adults in Dec. 2022 about AI in the workplace and its uses in hiring, monitoring, and evaluating workers.

Americans have mixed feelings about AI use in the workplace, with more believing it will significantly impact workers in general than themselves personally.

Other findings in brief:

  • đŸ€– Majorities oppose AI being used in making final hiring decisions, tracking workers' movements, and determining firing decisions.

  • đŸ§‘â€đŸ’Œ However, 47% think AI would do better than humans in evaluating job applicants, and a majority believe AI could help reduce racial and ethnic bias in worker evaluations.

  • 📉 More Americans think AI in the workplace will hurt workers than help them, while 32% believe the benefits and harms will be equally split.

  • đŸ’Œ 56% of Americans believe AI will have a significant impact on the U.S. economy over the next 20 years.

Customer service workers at a Fortune 500 software firm were 14% more productive on average when given access to generative AI tools, with the least-skilled workers benefiting the most, according to a study by researchers at Stanford University and the Massachusetts Institute of Technology.

While Scale AI conducted a survey on impact of generative AI and how companies are adopting AI in 2023 to improve their business. 65% of respondents increased their AI budgets or created an AI strategy for the first time due to the significant impact of generative AI in 2022.

A survey by KPMG U.S. reveals that most executives believe generative AI will have a higher impact on broader society than any other emerging technology. But they are not ready for immediate adoption due to the lack of technology, talent, and governance.

OpenAI

OpenAI filed a patent application in December 2022 due to a growing number of "myriad infringements and counterfeit apps" that use the acronym.

The USPTO stated that OpenAI neglected to pay an associated fee and provide "appropriate documentary evidence supporting the justification of special action." Hence, the patent was denied.

The company has also released a set of branding guidelines on app names. As an example, companies shouldn’t use MeowlyticsGPT. The name should rather be “Meowlytics powered by GPT.”

OpenAI has introduced new data management features in ChatGPT. Users can now disable chat history, which prevents conversations from being used to train models or appearing in the history sidebar. They are also introducing a ChatGPT Business subscription for professionals and enterprises seeking more control over their data. Additionally, a new export option has been added to make it easier to export chat data.

OpenAI has raised over $300 million at a valuation of $27 billion - $29 billion in a tender offer backed by investors such as Tiger Global, Sequoia Capital, Andreessen Horowitz, Thrive and K2 Global, according to documents seen by TechCrunch.

Government Regulations and Actions around AI

Italy has lifted the ban on the AI chatbot, ChatGPT, after its developers made changes to address data privacy concerns. OpenAI implemented changes that comply with several data privacy conditions, including increased transparency on its website, opt-out rights for users, and age verification for those accessing the site from Italy.

The European Union (EU) has reached an early agreement on its Artificial Intelligence (AI) Act, which would require companies to disclose any copyrighted material used to develop their generative AI tools. Show me your sources says the EU.

While in China - Companies must comply with China's strict censorship rules, ensuring the content of AI systems reflects "socialist core values." Chatbots must create truthful content and avoid materials that undermine "state power" or national unity, and respect intellectual property. Companies must register their algorithms, and the rules require a level of technical control over chatbots that Chinese tech companies have not achieved.

The U.S. Supreme Court rejected computer scientist Stephen Thaler's challenge to the U.S. Patent and Trademark Office's refusal to issue patents for inventions his artificial intelligence system created. The court upheld a lower court's ruling that patents can only be issued to human inventors, and not to AI systems.

Russia's Sberbank has launched its own AI chatbot called GigaChat to rival Microsoft-backed firm OpenAI's ChatGPT. GigaChat is currently in an invite-only testing mode and has the ability to converse more intelligently in Russian.

The UK government has pledged an initial ÂŁ100 million to establish a Foundation Model Taskforce which will develop AI to make the country "globally competitive," safer and more reliable, inspired by the COVID-19 vaccine unit. The taskforce will report directly to both the Prime Minister and Technology Secretary.

Yokosuka, a city in Japan, has become the first municipality in the country to use ChatGPT, an AI-powered chatbot, in its municipal offices for a one-month trial. This has been necessitated by Japan’s changing demographic

The US House of Representatives’ digital service has obtained 40 licenses for OpenAI’s ChatGPT Plus, which were distributed earlier this month, to test the technology internally. The new AI tool is expected to be used for generating constituent response drafts, press documents, summarizing speeches, drafting policy papers or bills, creating new logos, and more. — God help us all.

Palantir, Coco Cola and Others

Palantir has showcased its latest offering, the Palantir Artificial Intelligence Platform (AIP), designed to integrate large language models (LLMs) into privately-operated networks. In a demo video, the AIP is shown being used on the modern battlefield, where it helps estimate the enemy's composition and capabilities by launching a Reaper drone on a reconnaissance mission, suggesting appropriate responses given the discovery of an armored element, and even organizing the jamming of the enemy's communications.

The AIP's "industry-leading guardrails" will prevent the system from taking unauthorized actions, but a "human in the loop" to prevent such actions does exist in Palantir's scenario.

Coca-Cola has announced that it will be prioritizing AI in 2023. The company is collaborating with OpenAI and Bain & Company to build capabilities through cutting-edge AI to enhance marketing capabilities, business operations, and customer service.

It is also using AI to improve point-of-sale material creation in collaboration with its bottling partners. The company launched the Create Real Magic platform that invited consumers to use AI to generate original artwork with Coca-Cola’s iconic creative assets. The platform received 120,000 submissions of AI-generated images. In mid-July, Coca-Cola will bring 30 AI artists on an all-expenses-paid trip to the company’s headquarters in Atlanta.

Companies continue to leverage AI:

  • Intuit is hoping that AI will enable it to improve its offerings and position it as an "AI-driven expert platform". — Techcrunch

  • Walmart is using an AI-powered chatbot developed by Pactum AI to automate its vendor negotiations, allowing the AI to communicate with human sellers to close deals. — Bloomberg

  • Dropbox is laying off around 500 employees in an effort to build out its AI division and shift to early-stage product development. — TheVerge

  • Yelp rolling out a host of AI powered features — Venturebeats

  • TikTok has released an AI avatar generator tool that allows users to upload their photos and generate up to 30 avatars in five different styles to choose from. — Gizmodo

  • PwC plans to invest $1 billion over the next three years in generative artificial intelligence (AI) technology to automate aspects of its tax, audit and consulting services. — WSJ

  • Discourse AI is a new plugin that integrates Artificial Intelligence with Discourse, allowing community managers, moderators, and members to perform various tasks. — Discord

  • Binance has launched Binance Sensei, an AI-driven learning tool that is integrated with Binance Academy to help users access its vast array of educational resources in a more intuitive manner. — Binance

  • Apple is developing an AI-powered health coaching service called Quartz, as well as new technology for tracking emotions. The service will use data from the Apple Watch to create personalized coaching programs for users. — Bloomberg

AI-Generated Videos

First, we have a pizza ad generated entirely using AI. It is made of nightmare fuel.

And RNC released a AI generated video after Biden’s announced he will re-run for President in 2024:

Misc Stuff

That’s it folks. Thank you for reading and have a great week ahead.