Tech Brew Ride Home - Wed. 03/20 – Microsoft Hires DeepMind Co-Founder And Rocks The AI Community
Episode Date: March 20, 2024Microsoft hiring DeepMind co-founder Mustafa Suleyman to run their new AI division might sound like boring c-suite musical chairs, but it’s actually super interesting. Intel gets the first huge chec...k from the CHIPS Act. Stardew Valley is breaking gaming records. And the interesting startup that does AI music. Links: Microsoft Hires DeepMind Co-Founder Suleyman to Run Consumer AI (Bloomberg) Microsoft bets on start-ups to extend AI lead with hiring of Inflection chief (Financial Times) Intel to receive $8.5bn in US funding for high-end chip manufacturing (Financial Times) Stardew Valley’s 1.6 update smashes its Steam player record (The Verge) A ChatGPT for Music Is Here. Inside Suno, the Startup Changing Everything (Rolling Stone) Learn more about your ad choices. Visit megaphone.fm/adchoices
Transcript
Discussion (0)
On April 4th, 2023, around 2 in the morning, a man was found stabbed multiple times on a sidewalk in downtown San Francisco.
Hey, who did this to you?
What happened next turned the story into a political firestorm.
Reports have identified the victim as Bob Lee, the founder of Cash App.
From Bloomberg Podcasts, this is Foundering, the Killing of Bob Lee, beginning April 16.
Welcome to the Tech meme right home for Wednesday, March 20th, 2024. I'm Brian McCalla today. Microsoft hiring DeepMind co-founder Mustafa Suleiman to run their new AI division might sound like boring C-suite musical chairs, but it's actually super interesting. Intel gets the first huge check from the Chips Act. Star Do Valley is breaking gaming records and the interesting startup that does AI music. Here's what you miss today in the world of tech. Microsoft has hired DeepMind co-founder Mustafa Suleiman to run consumer.
A.I. at Microsoft, reporting directly to Satcha Nadella, and also brought on board most of the staff
from his Infliction AI startup, which had raised over a billion dollars in just the last year.
This is all part of Microsoft creating Microsoft AI, a division slash brand, focused on co-pilot
and other consumer AI products. Mustafa Suleiman will be its CEO and Karen Simonian, its chief
scientist. Quoting Bloomberg, Inflection, a rival of Microsoft.
key AI partner OpenAI is shifting to selling AI software to businesses, but will continue
operating its Pi Consumer Chatbot business for now. Inflection in June raised $1.3 billion
in one of the largest funding rounds of Silicon Valley's AI frenzy. That round valued the
startup at $4 billion, a person familiar with the matter said at the time. Microsoft board member
Reid Hoffman is a co-founder of the two-year-old startup alongside Suleiman and Simonian. And other
investors include Microsoft co-founder Bill Gates, former Google,
CEO Eric Schmidt and Invidia.
Inflexion's pie chatbot was designed to mimic human understanding of emotions and interact
with users in a supportive fashion.
While attracting considerable investor interests, including from Microsoft and a million
active daily users, the startup has not succeeded in finding an effective business model,
Suleiman said. In the meantime, Nadella asked him to move over to Microsoft.
The new hires also mark another significant step by Microsoft to bolster its in-house
AI capabilities and products outside of the relationship with OpenAI.
Last month, Microsoft invested $16 million in Mistral AI, a French rival to OpenAI.
Nadella on Monday told OpenAI, CEO Sam Altman, about Suleiman and his team joining Microsoft,
the company said, end quote.
Okay, so if it wasn't obvious why all of that is raising eyebrows all around the valley,
let me outline some of the interesting bits about this in no particular order.
So first off, this is one of the co-founders of Deep Mind, basically the OG,
of modern AI companies. So this is a superstar signing in almost the sports sense. Deep Mind is now
a division of Google, so there's that too, not exactly stealing a player from another team,
but effectively taking the possibility he would return to Google off the board. But then there's
the fact that this is not an acquisition. Microsoft did not acquire inflection. It simply
hired most of their staff. Like, remember, that was what was proposed when the whole Sam Altman
firing thing was going down. Before he went back to Open AI, the idea was, hey, Sam, come over here to
Microsoft, take over our AI efforts, and bring your team. So I guess they got the idea from that.
But like, how does that work? Inflection raised $1.3 billion in venture capital funding.
Reid Hoffman posted on LinkedIn that, quote, this agreement with Microsoft means that all of
inflections investors will have a good outcome today, and I anticipate good future upside, end quote.
But what does that mean? How are the investors getting paid?
What is the value of inflection if the best part of their team is no longer with the company,
as one has to presume Microsoft and Suleiman are taking all the top talent?
Is this just a clever way for Microsoft to do an acquisition and all but name?
Regulators are reportedly already circling Microsoft's tight relationship with open AI.
You'd have to imagine there would basically be no way Microsoft could have acquired a major
open AI competitor without regulators at least strongly looking at it.
So is this kosher? I mean, they have essentially taken out.
a major Open AI competitor, even if that competitor is ostensibly still operating. Then there's the fact
that Reid Hoffman, who was a co-founder of Inflection, is also on Microsoft's board of directors. Given that,
given Bill Gates was an investor in Infliction, was this deal made possible by close relationships with Microsoft ties?
I'm not suggesting anything nefarious is going on here. I'm just rattling off some of the questions being
asked around the valley. On that note, if you're Sam Altman and you get this news, how do you feel about your
relationship with Microsoft now. Like, why do you need Suleiman if you already have Altman? If you
already have de facto control of OpenAI, if not DeJure control? Is Microsoft moving away from OpenAI,
diversifying away, whether for regulatory reasons or maybe strategic or even personal reasons?
Like on the face of this, the biggest loser in this move would seem to be Open AI.
Does this suggest that the era of chatbots is already waning? Like, there are so many of them.
Claude, Rufus, Poe, Grock, Gemini, is this a commodity thing now? Is this like the era of search engines
before Google, when you had Excite and Alta Vista and Lycos and Asjeeves, all doing pretty much the same thing in a not very differentiated way?
Or is this a sign chatbots are a cul-de-sac? Before releasing Pi, Suleiman originally said inflection was going to develop a so-called AI chief of staff
that would manage various tasks without the need for prompting an interaction. He seemingly wasn't able to do that at a startup,
Is the idea that he can now attempt to make that leap beyond chatbots with Microsoft's unlimited funds and
resources? Finally, let me throw this idea out there that I've been hearing batted around a lot lately.
The idea that training AI models is such a specialized thing that people are wondering if a sort of
priesthood of AI whisperers is developing. I was just listening to a podcast where the AI investors
Nat Freeman and Daniel Gross were talking about how the whole process of training models,
refining models, fine-tuning them is so artisanal.
literally more art than science. In such a reality, you don't need entire companies. You just need
teams or individuals even. The value is in the literal priesthood, not something like IP or even
technology, because it doesn't really exist. The Genesequois is in the literal taste and
refinement of the people who know how to make the models dance in the way that an artist has a
specific style or refinement that others don't. Is this news proving that that may be the case?
Here's a quote from the Financial Times. There's not a list of more than 20 people who could do
Soleiman's new job, said Aaron Levy, chief of cloud company box, adding that Microsoft's lead
in generative AI had been driven by Nadella identifying the, quote, right people at the right
time, end quote. The U.S. has awarded Intel $8.5 billion in Chips Act funding and $11 billion in
loans to expand chipmaking operations in Arizona, Ohio, New Mexico, and Oregon. Intel plans,
to spend $100 billion over five years on such production expansion. Quoting the Financial Times.
U.S. President Joe Biden will travel to Intel's site in Chandler, Arizona on Wednesday to announce
the package, which will go towards building new facilities for the company in the southwestern state,
as well as in Ohio, New Mexico, and Oregon. The government funding for chip manufacturing,
which was passed by Congress in 2022, is part of Biden's sweeping agenda to revitalize
domestic manufacturing in areas ranging from clean energy to semiconductors and steel.
Intel has already committed to investing $100 billion in chip manufacturing over the next five years.
It had said it expected to further benefit from U.S. Treasury tax credits that would allow it to write off
up to 25% of that investment. The $8.5 billion will be distributed in tranches,
subject to Intel reaching certain milestones, senior White House officials said they expect the
funding to lead to 30,000 jobs in the chip sector. The officials said the funds for Intel
should start to arrive later this year once the agreement is finalized. In total, it is
likely to constitute the largest such grant made under the 2022 Chips and Science Act, which provided
for $52 billion in subsidies to help shift semiconductor manufacturing back to the U.S. amid geopolitical
tensions with China. Gina Romando, U.S. Commerce Secretary, told reporters the grant would put the U.S.
on track to meet its goal of ensuring that 20 percent of the world's most advanced chips are made
by the U.S. by the end of the decade. The vast majority of high-end semiconductors are at present
manufactured by TSM. The U.S. relies on, quote, a very small number of factories in Asia for all of our
most sophisticated chips, Romando said, which she described as an untenable situation from a U.S.
Economic and National Security perspective. Romando added that further grants under the Chips Act would
soon follow. TSM and Samsung, which also operate facilities in the U.S., are both awaiting
their own subsidies packages, end quote.
According to SteamDB, after releasing its massive 1.6 update, Stardew Valley hit 146,159
concurrent players on Steam, a new record on that platform and above Powellworld's stats this week.
Quoting the verge, Star Doe Valley's massive 1.6 update has been out on PC for less than 24 hours,
and it's already taken the game's popularity to staggering new heights.
The cozy farming simulator hit 146,159 concurrent players on
Steam, according to Steam DB data, smashing the game's previous record as players flocked to
enjoy new content, including a new farm, festivals, and pets. Stardue Valley's previous peak of
94,875 concurrent Steam Players was achieved back in January 2021, shortly after the 1.5 update was
released. By contrast, its new record is higher than any of the concurrent player stats reported
for Powell World this week, which is pretty impressive for a near-decade-old indie RPG. Developer
Eric concerned ape Barone had teased some of the new 1.6 update content on X in the run-up to the
patch being released yesterday, which likely helped to entice new and existing fans back to the game.
Some highlights in the update include a massive new Meadowlands farm optimized for animal rearing,
being able to move the farmhouse to a new location, new pets, including turtles and new cat-slash-dog
breeds, and support for multiple pets, and a whole bunch of new NPC dialogue.
You can also now drink mayonnaise.
things haven't been included in the patch notes, so there are a few surprises that will only be
discovered by playing the game. The Stardoo Valley 1.6 update is currently available on Steam and
G-O-G, a 1.6.1 patch to fix a few bugs, has also already been released. Barone hasn't announced
when the update will be ready for console and mobile devices, but says the Game Pass update is
incoming soon. Star Doe Valley is discounted to just $11.99 right now on Steam for anyone not
already swept in by the hype, end quote. Finally today, I've said video has been having an AI
moment, but let's not sleep on music. Rolling Stone takes a look inside generative AI music
startup, Suno, whose model can compose songs, including human vocals, using a text prompt, as Suno
aims to, quote, democratize making music. Quote, I'm just a soul trapped in the circuitry.
The voice singing those lyrics is raw and plaintive, dipping into blue notes.
A lone acoustic guitar chugs behind it punctuating the vocal phrases with tasteful runs.
But there's no human behind the voice, no hands on that guitar.
There is, in fact, no guitar.
In the space of 15 seconds, this credible, even-moving blues song was generated by the latest AI model from a startup named Suno.
All it took to summon it from the void was a simple text prompt.
Solo acoustic Mississippi Delta Blues about a sad AI.
To be maximally precise, the song is the work of two AI models in collaboration,
Suno's model creates all the music itself while calling on OpenAI's ChatGBT
BT to generate the lyrics and even a title, Soul of the Machine. Online, Suno's creations
are starting to generate reactions like, How the F is This Real? As this particular track
plays over a Sonos speaker in a conference room in Suno's temporary headquarters steps away
from the Harvard campus in Cambridge, Massachusetts, even some of the people behind the
technology are ever so slightly unnerved. There's some nervous laughter alongside murmurs of
Holy S and oh boy. It's mid-February and we're playing with their new model V3, which is still a
couple of weeks from public release. In this case, it took only three tries to get that startling
result. The first two were decent, but a simple tweak to my prompt. Co-founder Kenan Freyberg suggested
adding the word Mississippi resulted in something far more uncanny. Over the past year alone,
generative AI has made major strides in producing credible text, images via services like
mid-journey, and even video, particularly with OpenAI's new SORA tool. But audio and music in particular
has lagged. Suno appears to be cracking the code to AI music, and its founders' ambitions are nearly
limitless. They imagine a world of wildly democratized music-making. The most vocal of the co-founders,
Mikey Schulman, a boishly charming backpack-toting 37-year-old with a Harvard PhD in physics,
envisions a billion people worldwide, paying 10 bucks a month to create songs with Suno. The fact that
music listeners so vastly outnumber music makers at the moment is, quote, so lopsided, he argues,
seeing Suno as poised to fix that perceived imbalance. Most AI-generated art so far is at best
kits-a-a-a-the-hy-gum-fitting spacesuits that so many mid-jorney users seem intent on
generating. But soul of the machine feels like something different. The most powerful and unsettling
AI creation I've encountered in any medium. Its very existence feels like a fissure in reality,
at once awe-inspiring and vaguely unholy. And I keep thinking of the Arthur C. Clark quote that seems made for this generative AI era. Any sufficiently advanced technology is indistinguishable from magic. A few weeks after returning from Cambridge, I send the song off to living color guitarist Vernon Reed, who's been outspoken about the perils and possibilities of AI music. He notes his wonder, shock, horror, at the song's disturbing verisimilitude, the long-running dystopian ideal of separating difficult, messy, undesire.
and despise humanity from its creative output is at hand, he writes, pointing out the problematic
nature of an AI singing the blues, quote, an African-American idiom, deeply tied to historical
human trauma and enslavement, end quote. Suno is barely two years old. Co-founders Schulman,
Freiburg, and George Kukko, and Martin Camacho, all machine learning experts,
work together until 2022 at another Cambridge company, can show technologies which focused on
finding AI solutions to complex business problems. Shulman and Camacho,
are both musicians who used to jam together in their Ken Show days, at Ken Show the Forsome worked on a
transcription technology for capturing public company's earnings calls, a tricky task given the combination
of poor audio quality, abundant jargon, and various accents. Suno uses the same general
approach as large language models like ChatGPT, which breaks down human language into discrete segments
known as tokens, absorb its millions of usages, styles, and structures, and then reconstruct it
on demand. But audio, particularly music, is almost unfathomably more complex, which,
which is why, just last year, AI music experts told Rolling Stone that a service as capable as
Sunos might take years to arrive. Audio is not a discrete thing like words, Schroleman says.
It's a wave. It's a continuous signal. High-quality audio's sampling rate is generally 44 kHz
or 48 hertz, which means 48,000 tokens a second, he adds. That's a big problem, right?
And so you need to figure out how to kind of smush that down to something more reasonable.
How, though? Quote, a lot of work with lots of
heuristics, a lot of other kinds of tricks and models and stuff like that, I don't think we're
anywhere close to done. Eventually, Suna wants to find alternatives to the text-to-music interface,
adding more advanced and intuitive inputs, generating songs based on users' own singing is one idea.
OpenAI faces multiple lawsuits over chat, GPT's use of books, news articles, and other copyrighted
material in its vast corpus of training data, Suno's founders declined to reveal details of just what
data they're shoveling into their own model, other than the fact that its ability to generate
convincing human vocals comes in part because it's learning from recordings of speech in addition to music.
Naked speech will help you learn the characteristics of human voice that are difficult,
Schollman says, end quote.
When I am king, you will be first against the wall, with your opinion, which is of no consequence at all.
Figured after that last piece, the only appropriate song to quote lyrics from would be paranoid android.
Naturally, talk to you tomorrow.
