1 big thing: AI's great "pause" debate

2. Fantasy author takes on Audible's "poor" terms 

Photo illustration of Brandon Sanderson with Audible's logo and a stack of books.

3. Take note

4. After you Login 

They say you can't step in the same river twice. But, friends, here in the Bay Area, it has felt like we've been stuck in the same "atmospheric river" all season.

Today's Login is 1,175 words, a 4-minute read. 

Thanks to Peter Allen Clark for editing and Bryan McBournie for copy editing this newsletter.

Login

A smartphone with different colored buttons floating above its surface.

deepBlue

Taking you inside the AI revolution, and delivering scoops and insights on the technologies reshaping our lives.

Axios Login

Standard

1 big thing: Government begins to act on battery fires

Illustration of a series of batteries on fire creating an upward bar chart. 

2. Who's schooling Congress on AI 

Illustration of a US Capitol dome made out of binary code. 

4. After you Login

I've enjoyed steering the good ship Login this week! Ina will be back as usual Monday.

Today's Login is 1,254 words, a 5-minute read.

🚫 Situational awareness: The Italian privacy regulator temporarily banned ChatGPT, alleging privacy violations, Politico reports.

1 big thing: AI changes the software-making game

Illustration of binary code on a conveyor belt

2. Americans hooked on Chinese apps

Hello world! Ina's taking a break, and I will be your host for the Login week. 

Axios' second annual What's Next Summit is this Wednesday, March 29, in Washington, D.C.! You’ll hear from music producer Timbaland, head of YouTube Neal Mohan, White House Office of Science and Technology Policy director Arati Prabhakar and more on the issues and innovations shaping our future. Register to livestream the event here.

Today's newsletter is 1,202 words, a 5-minute read. 

💰 Situational awareness: First Citizens BancShares has agreed to buy the commercial banking business of Silicon Valley Bank, which recently collapsed and was seized by the U.S. government.

1 big thing: Congress zeroes in on content rules 

Illustration of hands with markers crossing out content on a laptop screen

2. Wellness lab offers parents TikTok counsel

Illustrated collage of TikTok’s logo and a hand holding a phone.

I know the days are getting longer, but this time of year, don't the months feel like they're lengthening too? 

🧩 Situational awareness: Chinese tech giant Alibaba announced today it will split into six core businesses that will be able to pursue independent fundraising and IPOs.

Today's Login is 1,232 words, a 5-minute read. 

1 big thing: SS Twitter keeps taking on more water 

Illustration of the Twitter bird surrounded by shark fins.

2. Microsoft brings generative AI to cyber with new tool

illustration of a screen of binary code with parts of it glowing in green to form a thought bubble

3. Take note 

Tune in today to Axios' second annual What's Next Summit, with events and speakers all day. Register here.

Today's Login is 1,069 words, a 4-minute read. 

FetchError

National

Advances in AI are imperiled by our failure to build the new technology with ethical care, says Rumman Chowdhury, scholar and co-founder of the AI non-profit Humane Intelligence.

Why it matters: While the media and policymakers are transfixed by warnings about long-term existential risks AI might pose, they are ignoring hard evidence of real harms happening today, Chowdhury told Axios' Ina Fried Tuesday at Axios' AI+ Summit in Washington, D.C.

What's happening: AI systems are already wrongly imprisoning people or denying them vital organs for transplants, she said. "These problems are existential, too, to the individuals impacted by them," she added, citing Vice President Kamala Harris' speech during the U.K.'s AI summit earlier this month.

"We've been sold this beautiful dream about how AI is going," Chowdhury said. But if the public sees AI being used to harm or discriminate, the technology will lose support and alienate potential users."

We can say we want AI to end global poverty, but since we have enough resources to do that today, the problem is not a lack of AI, but of human intention, she added.

One problem Chowdhury emphasized is the tech industry's continued failure to include diverse perspectives as it builds products for global use. AI has brought with it "a resurgence of bro culture," meaning "we're looking at tech 2017 all over again."

The outcome of the recent OpenAI battle, she noted, pushed two women out of the boardroom — and added former Treasury Secretary Larry Summers, who made headlines for minimizing women's aptitude for science when he was president of Harvard. 

Rumman Chowdhury: Ethical misfires undermine support for AI

The AI scholar also said the tech industry is once again building a new system without incorporating diverse viewpoints.

Rumman Chowdhury at Axios' AI+ Summit in Washington, D.C. Photo: Eric Lee/Axios

Artificial intelligence

Artificial Intelligence

Advances in AI are imperiled by our failure to build the new technology with ethical care, says <a class="gtmContentClick" data-vars-link-text="Rumman Chowdhury" data-vars-click-url="https://www.axios.com/2018/06/08/new-tools-to-make-ai-more-fair-1528464137" data-vars-content-id="4001eb11-c4f7-435e-907f-34ca1c138c5a" data-vars-headline="Rumman Chowdhury: Ethical misfires undermine support for AI" data-vars-event-category="story" data-vars-sub-category="story" data-vars-item="in_content_link" href="https://www.axios.com/2018/06/08/new-tools-to-make-ai-more-fair-1528464137" target="_self">Rumman Chowdhury</a>, scholar and co-founder of the AI non-profit Humane Intelligence.Why it matters: While the media and policymakers are transfixed by warnings about long-term existential risks AI might pose, they are ignoring hard evidence of real harms happening today, Chowdhury told Axios&#x27; Ina Fried Tuesday at Axios&#x27; AI+ Summit in Washington, D.C.

From the Valley to D.C., Big Tech players like Facebook, Google and Amazon are under more scrutiny than ever as new technology develops and privacy and antitrust concerns grow in lockstep with companies’ ambitions.

Technology

The lay of the Big Tech landscape has profoundly shifted in the year since OpenAI released ChatGPT on Nov. 30, 2022.

Why it matters: Tech's own innovation — rather than government or any other external force — has once again driven a massive transformation of the industry.

By contrast, five years of techlash — including hundreds of congressional hearings and bills, a barrage of lawsuits and precipitous drops in public trust — failed to unseat, or even slow, any of the industry's five dominating companies.

Of note: Two previous industry-led waves of new tech — the cryptocurrency-Web3 vision and the dream of building a metaverse — have yet to cross over from geeky niche enthusiasm to mass-market adoption. ChatGPT did so faster than any previous tech offering.

What's happening: Every one of tech's giants has begun to reorder its world around generative AI, which most of Silicon Valley now views as the biggest new-platform opportunity since the iPhone arrived in 2007.

Google quickly started putting AI-written summaries at the top of its search results, even as leaders and investors worried about how the change could undermine the search giant's ad-based profit machine.

Google also moved to transform DeepMind, which it acquired in 2014, from a research-oriented think tank into a product-focused operating unit.

The giant's internal progress on generative AI matched or exceeded OpenAI's before ChatGPT's release — but Google had approached public deployment more cautiously, concerned over potential risks and harms.

That ambivalence still looms over its strategy, despite reports of "code red" mobilization inside Google to meet the moment.

Microsoft — OpenAI's closest partner and biggest investor — rushed to wrap ChatGPT-style assistants, which it called copilots, into every corner of its enterprise-dominating office tools and operating systems.

Microsoft is now "the copilot company," CEO Satya Nadella told a developer conference this month. In the future, he said, "there will be a copilot for everyone and everything you do."

Meta is talking much less about building the metaverse — an initiative around which Mark Zuckerberg renamed the company just two years ago — and much more about the ways generative AI will drive growth in social media engagement and advertising.

Meta has built its AI strategy around releasing open source models that, it hopes, will prevent its rivals from controlling the AI platform the way Apple and Google owned the smartphone through their operating-system control.

Amazon hopes to parlay its Amazon Web Services subsidiary's dominance of cloud computing into a commanding position in generative AI.

With its Alexa voice assistant, the retail giant was a leader in the previous generation of conversational AI. But Amazon has yet to put a strong stamp on the consumer end of the new chatbot boom — though it, like Google, hopes the current AI boom will revitalize its voice-assistant business.

Apple remains Big Tech's most conspicuous laggard in the AI race ChatGPT set off.

While its Siri assistant introduced the public to voice-activated computing, and the firm is rumored to be at work on its own generative AI projects, Apple's biggest product rollout of 2023 involved its next-generation Vision Pro headset, rather than AI.

Yes, but: Apple remains tech's most valuable company.

Historically, Apple has waited to swoop in and seize markets only when it believes emerging tech and public demand have perfectly aligned — and its own innovations can provide margin-boosting added value.

How ChatGPT rewired the tech world

ChatGPT forced Microsoft, Google, Meta, Amazon and Apple to change plans and compete in new ways.

Illustration of a silver robot hand touching a surface and creating a glowing green ripple

Axios AI+

Sam Altman's return to OpenAI as CEO is a win for him, but not a total win — and even now, the game still has more innings.

Why it matters: The company that sparked the generative-AI boom isn't going to collapse in chaos or lose all its employees to competitors, but its future still carries big question marks.

What's next: The turmoil at OpenAI will calm down as a new three-man board takes over. But Altman isn't on that board — and Quora CEO Adam D'Angelo, part of the old board majority that fired Altman, is. 

An independent investigation into the events around Altman's firing could strengthen Altman's position if it concludes that he did no wrong. 

But if it finds a smoking gun somewhere in the former board's reasons for giving him the sack — which the board has failed repeatedly to explain since then — Altman could face more challenges.

The new board could move rapidly to expand its roster, including with representatives of Microsoft, the firm's largest investor and closest partner. 

But the board is unlikely to add Altman back until the investigation clears him.

That means ultimate authority over OpenAI's fate, which used to rest with six people, now rests with three: D'Angelo; new board chair Bret Taylor, former co-CEO of Salesforce; and former Treasury Secretary Lawrence Summers.

It's an all-white-male board deciding the future of a technology that's beset with concerns over bias, misinformation and deepening inequality. 

Zoom out: OpenAI's creation as a nonprofit was intended to help insure that the development of advanced AI would "benefit humanity." But since the release of ChatGPT a year ago, the company has been riding the tiger of a tech boom that could mint billions.

At first, the board battle looked like a last stand by the forces that aim to slow AI down for safety's sake against Silicon Valley's build-fast-and-break-things ethos.

In fact, sources familiar with the old board's thinking say the falling-out was more personal and involved a loss of trust in Altman.

The bottom line: There will be no pause in AI research and deployment. But OpenAI's near-implosion could reinforce efforts in Washington and abroad to regulate the AI industry more quickly.

Promises and commitments by CEOs don't look as reliable when the CEOs  can come and go this quickly. 

What's next for OpenAI with Sam Altman back in charge

Illustration of eye emojis with the OpenAI logo in the pupil

OpenAI

Sam Altman&#x27;s <a class="gtmContentClick" data-vars-link-text="return to OpenAI" data-vars-click-url="https://www.axios.com/2023/11/22/sam-altman-return-open-ai" data-vars-content-id="8d27c3e7-8c7a-4144-a5d9-1577b6e0ccb7" data-vars-headline="What&#x27;s next for OpenAI with Sam Altman back in charge" data-vars-event-category="story" data-vars-sub-category="story" data-vars-item="in_content_link" href="https://www.axios.com/2023/11/22/sam-altman-return-open-ai" target="_self">return to OpenAI</a> as CEO is a win for him, but not a total win — and even now, the game still has more innings.Why it matters: The company that sparked the <a class="gtmContentClick" data-vars-link-text="generative-AI boom" data-vars-click-url="https://www.axios.com/2023/03/23/generative-ai-chatgpt-capabilities" data-vars-content-id="8d27c3e7-8c7a-4144-a5d9-1577b6e0ccb7" data-vars-headline="What&#x27;s next for OpenAI with Sam Altman back in charge" data-vars-event-category="story" data-vars-sub-category="story" data-vars-item="in_content_link" href="https://www.axios.com/2023/03/23/generative-ai-chatgpt-capabilities" target="_self">generative-AI boom</a> isn&#x27;t going to collapse in chaos or <a class="gtmContentClick" data-vars-link-text="lose all its employees" data-vars-click-url="https://www.axios.com/2023/11/20/openai-staff-letter-board-resign-sam-altman" data-vars-content-id="8d27c3e7-8c7a-4144-a5d9-1577b6e0ccb7" data-vars-headline="What&#x27;s next for OpenAI with Sam Altman back in charge" data-vars-event-category="story" data-vars-sub-category="story" data-vars-item="in_content_link" href="https://www.axios.com/2023/11/20/openai-staff-letter-board-resign-sam-altman" target="_self">lose all its employees</a> to competitors, but its future still carries big question marks.

Latest news on Wall Street, Corporate America and financial regulation.

Business

Guardrails AI companies add to their products to prevent them from causing harm "aren't enough" to control AI capabilities that could endanger humanity within five to ten years, former Google CEO Eric Schmidt told Axios' Mike Allen on Tuesday.

The big picture: Interviewed at Axios' AI+ Summit in Washington, D.C., Schmidt compared the development of AI to the introduction of nuclear weapons at the end of the Second World War.

"After Nagasaki and Hiroshima, it took 18 years to get to a treaty over test bans and things like that. We don't have that kind of time today," Schmidt said. 

The danger, he said, arrives at "the point at which the computer can start to make its own decisions to do things" — when, say, such a system discovers access to weapons, and we can't be certain the system will tell us the truth.

Two years ago, that moment was expected to be 20 years off. Today, Schmidt said, some experts think it's only two to four years away. 

What's next: Schmidt argued that the best solution is to create a global body akin to the Intergovernmental Panel on Climate Change (IPCC) to "feed accurate information to policymakers" so that they understand the urgency and can take action.

Of note: Schmidt said he's optimistic that AI will offer wide benefit to vast human populations: "I defy you to argue that an AI doctor or an AI tutor is a negative. It's got to be good for the world."

On the OpenAI boardroom drama: "It's pretty simple. The board fires Sam [Altman]. Sam fires the board." Once the OpenAI staff showed its loyalty to Altman with a mass open letter, Schmidt said, the outcome was inevitable: "How much more feedback do you need from your 360?" 

Go deeper: 3 reasons why the OpenAI meltdown matters

3 reasons why the OpenAI meltdown matters

Former Google CEO: Companies' AI guardrails "aren't enough" to prevent harm

At Axios' AI+ Summit in Washington, Schmidt advocated for a global advisory board.

Eric Schmidt at Axios' AI+ Summit. Photo: Eric Lee/Axios

Guardrails AI companies add to their products to prevent them from causing harm &quot;aren&#x27;t enough&quot; to control AI capabilities that could <a class="gtmContentClick" data-vars-link-text="endanger humanity" data-vars-click-url="https://www.axios.com/2023/05/02/ai-chatgpt-disasters-scenarios-harms-geoffrey-hinton" data-vars-content-id="d08eb826-9538-4800-bff1-0e0f7f0dc77c" data-vars-headline="Former Google CEO: Companies&#x27; AI guardrails &quot;aren&#x27;t enough&quot; to prevent harm" data-vars-event-category="story" data-vars-sub-category="story" data-vars-item="in_content_link" href="https://www.axios.com/2023/05/02/ai-chatgpt-disasters-scenarios-harms-geoffrey-hinton" target="_self">endanger humanity</a> within five to ten years, former Google CEO Eric Schmidt told Axios&#x27; Mike Allen on Tuesday.The big picture: Interviewed at <a class="gtmContentClick" data-vars-link-text="Axios&#x27; AI+ Summit" data-vars-click-url="https://www.axios.com/2023/11/17/axios-event-watch-ai-summit-washington-dc" data-vars-content-id="d08eb826-9538-4800-bff1-0e0f7f0dc77c" data-vars-headline="Former Google CEO: Companies&#x27; AI guardrails &quot;aren&#x27;t enough&quot; to prevent harm" data-vars-event-category="story" data-vars-sub-category="story" data-vars-item="in_content_link" href="https://www.axios.com/2023/11/17/axios-event-watch-ai-summit-washington-dc" target="_self">Axios&#x27; AI+ Summit</a> in Washington, D.C., Schmidt compared the development of AI to the introduction of nuclear weapons at the end of the Second World War.

New technologies always arrive bearing new vocabularies. Here's a guide to the lingo of artificial intelligence.

Artificial intelligence (AI): A term coined in the 1950s by John McCarthy to describe efforts to develop machines that could reason and solve problems in human-like ways. Now widely applied to any software that can identify patterns in data.

Machine learning and neural networks: An approach that took off in the 2000s and became the foundation of today's AI. Instead of being programmed in exhaustive detail to "know" bodies of knowledge, these systems feed on hoards of data and gradually refine their ability to make sense of the information, sometimes guided by human feedback.

Artificial general intelligence (AGI): A label computer scientists apply to the still-unachieved goal of creating AI that can reason and learn in broad ways;  apply those skills to new realms it hasn't encountered before; and grow in unpredictable ways.

Autonomy: The capacity of AI to act on its own to achieve a goal without specific human direction at every step — in the physical world (self-driving cars), in virtual environments (non-player characters in games) or on computer networks (personal assistants). 

Self-awareness or sentience: The ability of AI to "know" that it exists and has continuity in time.

Alignment: The ways in which the goals of the people building AI do, or don't, match the goals of the systems they are creating. 

Generative AI (or genAI): Machine-learning based AI that trains on sets of real-world data — most commonly images and text — to learn to predict or "generate" the next word or pixel in a sequence, creating the capacity to "write" new texts and "make" new images. 

Transformers: A machine-learning programming approach introduced by Google researchers in 2017 that turbocharged the ability to create generative AI. (Not to be confused with Hollywood's robot/cars.)

Natural language: How AI researchers describe the languages humans speak. "Natural language processing" means making human language intelligible to machines. 

Large language model (LLM): An AI program with a mathematical map — across a very large number of dimensions — of the relationships among a large number of words, usually broken down into tokens.

Token: Technical term for the unit generative AI models use to create their mathematical maps. People use words and sentences, but AI breaks them down into more uniform-sized tokens — chiefly for reasons of computing efficiency.

Training data: The data initially provided to an AI model for it to create its map of relationships.

Supervised and unsupervised training: If the training data has been labeled by humans in advance, giving the AI signposts and hints for how to organize it, the training is considered supervised. In unsupervised training, the model is simply turned loose on raw data, and the model gradually draws connections among tokens based on proximity.

Generative pre-trained transformer (GPT): A particular kind of LLM design, introduced by OpenAI, that uses a hybrid training approach, with an initial "pre-training" that is unsupervised and then a supervised "fine-tuning" phase. 

Compute: AI industry shorthand for the costly computer time required for all this training. The larger the model, the more "cycles" needed — and the greater the value of the most advanced, speediest processors (chips).

ChatGPT: OpenAI's runaway hit service that brought LLMs and generative AI into the mainstream by packaging it as a simple chatbot that could converse with users, compose stories and poems, answer questions and more.

Chatbot: Any program that can take on one half of a conversation with a user. Chatbots have been around since the 1960s, when Eliza — a simple chatbot that mimicked a therapist — first showed people were eager to personify computers. 

Turing test: A thought experiment proposed by computing pioneer Alan Turing in 1950. The Turing test measured whether a computer program could fool a human user via a blind onscreen chat into believing it was human too. 

Prompt: The text users enter to describe what they want from a generative AI program like ChatGPT or its image-making equivalents, like OpenAI's Dall-E.

Multi-modal: An AI system that can take input and produce output across different categories of media, typically text, images, audio and video.

Prompt engineering: The practice, often hit or miss, of trying to optimize prompts to produce exactly the output a user desires. 

Prompt injection:  Like prompt engineering, but with the goal of defeating restraints AI makers have built in to limit the production of potentially harmful content — for instance, instructions for bomb-makers. 

Context window: The short-term memory of a generative AI. The larger the context window, the more information you can "feed" the AI along with a prompt, giving it key ingredients upfront or new data that it might not have access to on the open internet.

Hallucination: An answer provided by generative AI that sounds plausible but is made up and incorrect. If the program does not have good information to go on, it will still try to answer a question by guessing "next words" that seem to fit. 

Deepfake: Any images, photo or video produced by AI tools designed to fool people into thinking it's real. 

Frontier model: An AI model that pushes the limits of what the most advanced programs can do today, and therefore pose the most potential risk.  

Existential risk: Danger that a powerful AI system might threaten humanity's future because of malfunction, unintended consequences or failures of alignment. 

Explainable AI: Systems that make it possible for programmers and researchers to trace how they arrived at any particular output or response. Many AI systems today are "black boxes" built without the kind of tooling that allows for such explanations. 

Go deeper: How AI works, in plain English

All the AI terms you need to know

From hallucinations to transformers  and AGI to LLMs, here's a crib sheet to all the AI lingo you're hearing.

Illustration of a large dictionary with AI embossed on the front

New technologies always arrive bearing new vocabularies. Here&#x27;s a guide to the lingo of <a class="gtmContentClick" data-vars-link-text="artificial intelligence" data-vars-click-url="https://www.axios.com/2023/02/17/chatgpt-ai-next-platform-tech" data-vars-content-id="fb79cea4-6639-4441-a58c-baf6a4b60b2b" data-vars-headline="All the AI terms you need to know" data-vars-event-category="story" data-vars-sub-category="story" data-vars-item="in_content_link" href="https://www.axios.com/2023/02/17/chatgpt-ai-next-platform-tech" target="_self">artificial intelligence</a>.Artificial intelligence (AI): A term coined in the 1950s by John McCarthy to describe efforts to develop machines that could reason and solve problems in human-like ways. Now widely applied to any software that can identify patterns in data.

Scott Rosenberg

Stories

Rumman Chowdhury: Ethical misfires undermine support for AI

Former Google CEO: Companies' AI guardrails "aren't enough" to prevent harm

How ChatGPT rewired the tech world

All the AI terms you need to know

What's next for OpenAI with Sam Altman back in charge

Newsletter editions

Axios Login: Lithium battery scares

Axios Login: AI's "pause" debate

Axios Login: Twitter's slow woes

Axios Login: Congress' tech obsessions

Axios Login: Software game-changer