Technology

➚ MOVE FAST: Data churn. If you don’t think AI models are going to keep getting bigger, just read this job posting from Microsoft. It’s hiring an expert to power its vast data centers with small, modular nuclear reactors.

➘ BREAK THINGS: Cash burn. OpenAI is already thinking about working on a hardware device, according to The Information. That would add to its growing expenses, which a paper valuation of up to $90 billion won’t solve.

Price of a monthly subscription to Tinder Select, the dating app’s new service designed for “highly engaged” users. For nearly $6,000 a year, loyal Tinder fans can message people they haven’t matched with and ensure their profiles are seen by “the most sought-after” potential dates, among other perks. The eye-popping price makes Tinder’s other subscription offerings, which start at $24.99 a month, look like a bargain.

Last week, Amazon unveiled a slew of products revamped with generative AI models running everything from the Echo to the Fire Stick. Amazon head scientist Rohit Prasad is leading the effort to bring generative AI into people’s homes.

Q: You stepped Alexa up to this LLM-style, more conversational experience. Does it hallucinate like some of the other foundation models?

A: LLMs are a waypoint towards generalized intelligence. They aren’t the endpoint. What you are seeing is still early days. All LLMs still hallucinate a bit. What we have done is taken extreme steps on how to ground them. If you said, ‘Alexa, it’s hot in here,’ a browser to an LLM-based system will say, ‘go to the beach.’ But if you asked Alexa on your home device, it should be grounded by knowing you have a thermostat connected to it. It should ask, ‘Do you want me to lower the temperature by five degrees?’

That’s the kind of important grounding that we have done. In addition, hallucinations are common because these are built on a token predictor. So you have to do a lot of fine tuning, as well as aligning so that it doesn’t hallucinate. Over the years of working on Alexa in the consumer domain, we have learned a lot on how to align these models.

Q: Is the model a continuation of the Alexa model or is it an entirely new model?

A: This is a new model we have built. But we’ve had years of experience with encoder-decoder models, which we were using for Alexa to learn new languages, new domains. And then we also had a visual language model, so that you can ask about product features. And some of the visual language models are also useful for things like visual processing. So we have had an immense amount of experience with that.

But this is a new model. Very large. And then it’s fusing in real-time devices and services, your personal context of what you watch, what you listen to, what’s your favorite teams, are you a vegetarian. And of course, how to align these models the right way, especially when you connect it to the real world.

Q: One of the big ideas in generative AI is the agent concept — an AI that can shop for you or buy your airline tickets. But there’s also some disillusionment that these agents are not going to work as well as we thought. You demonstrated one way it can work. Do you see this as a service within Amazon’s ecosystem?

A: Alexa is that super intelligence that works on your behalf. Today, it already controls your home. It assists you on your tasks. It’s being used by two-year-olds to 100-year-olds. It doesn’t take huge imagination to say, ‘if I’m running out of milk, you can order it.’

Then there will be many vertical agents. They’re specialized at a certain task. So if you have a law firm, you have an agent that you can build for your work. If you are in the travel industry, you may have a travel agent. But there will also be these super agents or the super AI like Alexa. Both are possible now.

My friend Eric Newcomer writes the startups and venture capital newsletter Newcomer. On November 15, he’s hosting an artificial intelligence conference in San Francisco with the AI gaming company Volley. Reid Hoffman, Vinod Khosla, Mustafa Suleyman, and many other top AI founders, investors, and executives will be speaking. Check out the Cerebral Valley AI Summit here.

Juozas Kaziukėnas is the founder and CEO of the e-commerce market research firm Marketplace Pulse.

James Anderson leads government innovation programs at Bloomberg Philanthropies, which organizes the annual CityLab conference. This week, the organization announced that Helsinki Mayor Juhana Vartiainen and Maryland Governor Wes Moore will be among the speakers at the October event in Washington, D.C. Louise spoke to Anderson about how tech is changing the future of cities.

Q: What are some biggest innovations that you anticipate will change city life over the next few years?

A: There are three shifts that are transforming the way local governments work and, as a result, have tremendous potential to deliver impact for residents and progress in cities. The first is a move from analog to digital and AI, which can help cities break free of the organizational constraints and operational silos that drag public institutions down. The second is an identity shift — from cities as service providers to cities as platforms — that invite and support the whole of their communities to contribute to new solutions that meet local ambitions and achieve civic goals.

The third shift is around participation: Local governments are increasingly generating policy with, rather than just for, residents. By that, I mean with youth or under-invested communities, instead of for them. It’s a powerful change that, as we are beginning to see in many cities’ climate efforts, will strengthen their legitimacy and their policies’ durability.

Q: Which technologies are you less excited about or think will wind up having little impact on cities?

A: Emerging technologies like generative AI do present extraordinary opportunities to reimagine public services, create efficiencies in city halls, and improve results. But getting there requires, first, building fluency around this new technology among mayors and key city officials. That upfront skills-building is fundamental to move governments beyond the hype, and to where cities are leveraging generative AI as a force to propel communities forward.

Q: Silicon Valley elites are currently trying to build their own city on farmland northeast of San Francisco. What do you think about these kinds of initiatives?

A: We have to summon the will and the skills to make sure our existing cities work better for everybody in them. This requires investing differently in local governments and other public institutions, not by expanding services or creating expensive new programs, but by strengthening local government talent and bolstering cities’ creative, collaborative, and analytical capacity to push solutions over the finish line.

• OpenAI announced that ChatGPT now has voice and audio capabilities. “Interestingly I felt heard & warm. Never tried therapy before but this is probably it?” an OpenAI employee tweeted after having an audio conversation with the chatbot about her personal life.
• Getty Images is partnering with Nvidia to launch an AI image-generation tool trained on its own vast photo library. Earlier this year, Getty sued Stability AI for allegedly using its images to train Stable Diffusion without properly licensing them first.
• YouTube creators can now feature ads in videos about controversial issues, including sexual and domestic abuse, eating disorders, and abortions — as long as they don’t go into graphic detail. YouTube said it wanted to ensure that creators weren’t disincentivized from making content that could be helpful to people.

• A large cache of Iranian government correspondence and emails offers a rare window into how Tehran works behind the scenes to bolster its image abroad.
• Why did California Governor Gavin Newsom veto a high-profile trans bill?
• All signs are still pointing to a U.S. government shutdown — but Republican moderates might team up with Democrats to keep it brief.