www.thedeepview.co Open in urlscan Pro
2606:4700::6812:4528  Public Scan

Form analysis
1 forms found in the DOM

POST /create

<form class="group w-full rounded-wt bg-transparent" action="/create" method="post" data-sparkloop-form-id="1"><input hidden="" value="" name="ref"><input hidden="" value="" name="bhba"><input hidden="" value="8c2a1b82-c99f-44f9-a71d-ecc670cc7f7d"
    name="visit_token"><input type="hidden" value="" name="cf-turnstile-response"><input type="hidden" name="redirect_path" value="/subscribe?recommendations=true&amp;email="><input type="hidden" name="sent_from_orchid" value="true"><input
    type="hidden" name="fallback_path" value="/p/microsoft-makes-bill-gates-ai-dreams-come-true"><input type="hidden" name="is_recaptcha_enabled" value="false"><input type="hidden" name="double_opt" value="false"><input type="hidden"
    name="trigger_redirect" value="true"><input hidden="" name="subscribe_error_message" value="Oops, something went wrong."><input hidden="" name="subscribe_success_message" value="Subscribed!">
  <div class="flex flex-col">
    <div style="background-color:#F9FAFB;border:2px solid #0500ff" class="flex w-full flex-col items-center sm:flex-row overflow-hidden p-1 rounded-lg">
      <div class="flex w-full items-center" style="background-color:#F9FAFB">
        <div class="pl-2" style="color:#0500ff"><svg xmlns="http://www.w3.org/2000/svg" fill="none" viewBox="0 0 24 24" stroke-width="1.5" stroke="currentColor" aria-hidden="true" data-slot="icon" class="h-6 w-6">
            <path stroke-linecap="round" stroke-linejoin="round"
              d="M21.75 6.75v10.5a2.25 2.25 0 0 1-2.25 2.25h-15a2.25 2.25 0 0 1-2.25-2.25V6.75m19.5 0A2.25 2.25 0 0 0 19.5 4.5h-15a2.25 2.25 0 0 0-2.25 2.25m19.5 0v.243a2.25 2.25 0 0 1-1.07 1.916l-7.5 4.615a2.25 2.25 0 0 1-2.36 0L3.32 8.91a2.25 2.25 0 0 1-1.07-1.916V6.75">
            </path>
          </svg></div><input type="email" autocomplete="email" required="" style="font-family:Open Sans;color:#111827"
          class="wt-button-font z-10 w-full border-none bg-transparent placeholder-shown:text-ellipsis text-lg focus:text-lg active:text-lg sm:text-lg" placeholder="Enter Your Email" name="email">
      </div><input type="submit" style="background-color:#0500ff;color:#FFFFFF;font-family:Open Sans" class="cursor-pointer px-5 py-3 font-semibold w-full sm:w-auto text-lg focus:text-lg active:text-lg sm:text-lg rounded-md" value="Subscribe">
    </div>
  </div>
</form>

Text Content

The Deep View

LoginJoin Free


0

 * The Deep View
 * Posts
 * ⚙️ Microsoft makes Bill Gates’ AI dreams come true


⚙️ MICROSOFT MAKES BILL GATES’ AI DREAMS COME TRUE

Ian Krietzberg
September 17, 2024



Good morning. Sean Astin (AKA, Samwise Gamgee, from Lord of the Rings) came out
alongside SAG-AFTRA in support of SB-1047, urging Gov. Gavin Newsom to sign the
bill.

As someone pointed out on Twitter, o1 (OpenAI’s new model) spelled backward is
One Ring…

— Ian Krietzberg, Editor-in-Chief, The Deep View

In today’s newsletter:

 * 🪐 AI for Good: Exoplanet discovery

 * 🤖 Google DeepMind makes progress in robotic dexterity

 * 🧑‍🍳 Chipotle adds robots to its staff

 * 💻 Microsoft makes Bill Gates’ AI dreams come true


AI FOR GOOD: EXOPLANET DISCOVERY 

Source: NASA

For four years, NASA’s Kepler Mission observed more than 200,000 stars in a deep
search for signs of distant planets in orbit around stars. 

In 2021, NASA deployed an AI algorithm — ExoMiner — to parse through the years
of data gathered by the mission to identify exoplanets.

The algorithm added 301 validated exoplanets to NASA’s total exoplanet tally. 

The details: The deep neural network is able to distinguish a real (or
validated) exoplanet from any number of false positives. The process of
validation involves statistical measures of certain planet-specific features. 

 * The 301 planets identified by ExoMiner were unable to be validated by
   scientists before the introduction of the algorithm. 

 * Perhaps the most important quality of ExoMiner is that, according to NASA, it
   “isn't a black box — there is no mystery as to why it decides something is a
   planet or not.”

Why it matters: This level of guaranteed explainability makes the system
reliable and trustworthy, allowing it be leveraged by scientists without
restraint. Now that it’s validated on the Kepler mission, the same system will
be used on other missions, all to aid the search for more exoplanets.


HOW TO DEMONSTRATE SECURE AI PRACTICES WITH ISO 42001



AI is everywhere, but customers need to know that you're using it safely and
responsibly. The ISO 42001 standard helps companies demonstrate their AI
security practices in a verifiable way.

Join Vanta and A-LIGN for a Coffee and Compliance session on ISO 42001—what it
is, what types of organizations need it, and how it works. The discussion also
covers practical strategies and best practices for successfully integrating ISO
42001 into your organization.

Watch the recorded session.


GOOGLE DEEPMIND MAKES PROGRESS TOWARD ACHIEVING ROBOTIC DEXTERITY

Source: Google DeepMind

Google DeepMind last week published two new papers on achieving greater levels
of robotic dexterity, something that might enable robots to become actually
useful and usable in daily life. 

The details: Where most advanced robots can only pick up and place objects with
a single arm, DeepMind’s research — ALOHA Unleashed — was able to crack bi-arm
manipulation. 

 * The two-armed robotic system was trained to mimic tasks that were first
   remotely demonstrated by the researchers. The robot was able to tie a
   shoelace, hang a shirt and clean a kitchen. 

 * DeepMind at the same time unveiled DemoStart, a reinforcement learning
   algorithm that helps robots “learn” these highly dexterous tasks in
   simulation. 

“One day, AI robots will help people with all kinds of tasks at home, in the
workplace and more,” DeepMind said. “Dexterity research, including the efficient
and general learning approaches we’ve described today, will help make that
future possible.”

Despite recent advances in robotics, that end goal of developing a C-3PO-like
butler robot, brimming with common sense, sarcasm and the ability to interact
with the physical world, as DeepMind notes, is far away. The challenges faced by
robotics go beyond those faced by AI (reasoning and reliability), and into
challenges around sensors and mechanics and physics. 


TRANSFORM YOUR HIRING GAME WITH ASHBY



For a hectic startup, hiring can be a messy process. Ashby makes it so much
easier. (That’s why writing this particular slot was so much fun — at The Deep
View we use Ashby and love it).

Their interface is intuitive, and applicants are easily organized, making
reviews quicker, more targeted, and as painless as possible.

With Ashby, our scheduling is almost completely automated, and as it turns out,
their new AI-assisted application review actually enables us to deliver a much
more human experience. Having all our workflows in a single tool helps us
deliver a better candidate experience with better data and more accuracy.

Plus, all of Ashby’s AI-backed features are reliable, secure and trustworthy (it
checks all the boxes).

If you’re hiring — but want to handle that process without your usual migraines
— switch to Ashby.

 * Building AI systems? Whether you're a beginner or looking to build confidence
   in launching to production, Vellum’s new webinar series is for you. Sign up
   today, space is limited.*

 * Tired of receiving compliance notices from state tax agencies? Automate
   payroll, compliance, and benefits management with Warp. Get started today and
   get a $1,000 gift card.*



 * The Air Force is researching an AI to help drones adapt to their environments
   (404 Media).

 * Data center emissions probably 662% higher than big tech claims (The
   Guardian).

 * TikTok faces tough questions from court over challenge to US law (Reuters).

 * How Google got away with charging publishers more than anyone else (The
   Verge).

 * Biden administration awards Intel up to $3 billion under the Chips Act
   (CNBC).

If you want to get in front of an audience of 200,000+ developers, business
leaders and tech enthusiasts, get in touch with us here.









… SPEAKING OF ROBOTS, CHIPOTLE IS TRYING SOME OUT

Source: Chipotle

Chipotle on Monday announced the integration of self-described “cobots,” which
the fast food chain describes as “collaborative robots,” in two California-based
locations. 

The details: One store is getting an “Autocado,” a machine that cuts and scoops
avocados. The other is getting an “augmented makeline,” which automates part of
the bowl (and salad) production process. 

 * Chipotle said that the Autocado takes 26 seconds to fully remove the flesh
   from a given avocado. (Chipotle uses 129.5 million pounds of avocados every
   year).  

 * The makeline automatically dispenses rice, corn, lettuce and other
   ingredients into a bowl beneath the counter, which employees can then
   customize. 

Chipotle said it had invested in Vebu (the company behind the Autocado) and
Hyphen (the company behind the makeline) through its $100 million Cultivate Next
venture fund. 

The context: Fast food chains have been trying to integrate generative AI and
robotics for a little while now. McDonald’s recently scrapped its AI drive-thru
ordering system after it kept screwing up; at around the same time, Taco Bell
said it plans to expand AI drive-thru ordering to hundreds of locations by the
end of the year.

Such automation also highlights concerns of job loss. Chipotle’s cobotic testing
ground is in California, where fast food chains are required to pay workers a
minimum of $20 an hour. 

I asked Chipotle if it plans to use the new cobots to hire fewer workers, or
otherwise trim its staff. A company spokesperson told me that “Autocado and the
Augmented Makeline will not eliminate any jobs. Crew members will have a cobotic
relationship with the devices in our restaurants.”


MICROSOFT MAKES BILL GATES’ AI DREAMS COME TRUE 

Source: Bill Gates

Almost exactly a year ago, Microsoft launched its first version of Copilot, a
genAI-based Windows feature that the company pitched as a “digital companion for
your whole life.” 

Yesterday, Microsoft unveiled a second wave of Copilot. This time, the spotlight
is on something called Copilot “agents.” Microsoft said that customers will be
able to custom-build AI agents that can then carry out tasks across Microsoft’s
suite of apps as well as third-party websites. 

Microsoft said the agents can range from “simple, prompt-and-response agents to
agents that replace repetitive tasks to more advanced, fully autonomous agents.”

 * There’s been a lot of talk of AI agents, lately. Salesforce recently unveiled
   “Agentforce,” which offers basically the exact same thing.

 * Google is working on its own AI agents, and then, of course, there’s Apple
   and Amazon, which are working on overhauling Siri and Alexa, respectively,
   with generative AI.

The idea of a legitimately reliable (and trustworthy) agent has had Microsoft
co-founder Bill Gates excited for a while. In early 2023, Gates said that the
first company to develop a good digital assistant would win the AI race, since
an agent would negate the need for manual online searching and shopping.

The other updates: In addition to the agents, Microsoft announced Copilot Pages,
intended to serve as a hub of customer-specific data that Copilot can generate
content from. 

 * Microsoft said Pages will be collaborative across teams, enabling
   “human-to-AI-to-human collaboration.” 

 * Microsoft is also shipping improvements to Copilot in Excel, PowerPoint,
   Teams, Outlook, Word and OneDrive. The big Excel improvement involves a new
   ability for Copilot to generate code in Python directly in Excel. 

Microsoft said that its Copilot customer base has grown some 60%
quarter-over-quarter. Copilot costs $30 per user per month in addition to the
cost of an existing Microsoft 365 plan, which costs between $6 and $22 per user
per month. 

The Bubble: Microsoft — and the industry on the whole — need this to work. The
problem is, so far, things aren’t going as smoothly as the Big Tech players
would like you to believe.

There are a few issues inherent to the large language models (LLMs) that power
Microsoft’s generative AI technology, including Copilot. A major one involves
hallucination, genAI’s propensity to generate confidently incorrect output;
another involves explainability. 

 * Current LLMs have been described as black boxes; we don’t know exactly how
   they work, or why they generate a certain output in response to a certain
   query. 

 * This lack of explainability raises issues of trust, reliability,
   discrimination, fairness and over-reliance that, in the enterprise (and
   everywhere else, frankly) are cause for concern. 

Despite the rosy picture of time-savings and productivity enhancements that
Microsoft painted in its post about the new wave of Copilot, The Information
recently reported that some Microsoft Copilot customers are pausing their
rollout of the feature due to annoying buys, high costs and unclear value. 



Right now, the AI ecosystem is sustaining itself in a strange, almost
closed-loop cycle; Nvidia sells super expensive AI chips, Big Tech buys billions
of dollars worth of those chips to build AI, and everyone’s stock prices soar. 

But if the companies that are using those Nvidia chips can’t generate legitimate
returns, they may have to stop buying new chips. If that happens, it could
signal the bursting of the AI Bubble, and everyone’s share prices will likely
plummet.

This, in essence, is the $600 billion gap in the industry — $600 billion in
spend, a few billion in return.

If reliability cannot be guaranteed, I’m not sure if this new round of agents
will be enough to crack the enterprise at a scale that would justify the ongoing
size of the investment. 

But I suppose we’ll see what happens as it rolls out.





WHICH IMAGE IS REAL?

 * ⬆️ Image 1
 * ⬇️ Image 2

Login or Subscribe to participate in polls.






🤔 YOUR THOUGHT PROCESS:

Due to technical issues with our CMS, I have no idea how you fared on this one.
The voting numbers are obscured. But, if you selected Image 2 (Left), congrats,
you won! If you selected Image 1 … swing and a miss.

SELECTED IMAGE 1 (RIGHT):

 * “Oh man! That was such a good, natural fence compared to the stamped out
   thing in #2. I think your AI has started living in the woods.”

If my AI ‘lived’ anywhere, it would absolutely be the woods. Have I mentioned
how much I like trees?

SELECTED IMAGE 2 (LEFT):

 * “I could only verify the shadows cast on this image, and I didn’t think AI
   would get the telephone/power line right.”


💭 A POLL BEFORE YOU GO

Thanks for reading today’s edition of The Deep View!

We’ll see you in the next one.


HERE’S YOUR VIEW ON AI GOVERNANCE:

Again, I have no idea how you voted on this one. All I have are your thoughts,
which I always appreciate.

Something else:

 * “Registration at least may be a good place to start with an assessment of
   regulatory need before wholesale jump to regulation.”

Developers should NOT be required to share models with governments:

 * “No, but that said, they should make their products identifiable as well as
   be held responsible for any harm they might be held responsible for.”


DO YOU USE COPILOT? HOW DO YOU LIKE IT?

 * It's amazing. As good as advertised
 * It's okay - not really a time-saver
 * I hate it. Waste of money
 * Will never use Copilot
 * Something else

Login or Subscribe to participate in polls.



KEEP READING


⚙️ REPORT: AI FOR HEALTHCARE PREDICTION NEEDS TO SLOW DOWN








⚙️ THE COMPLICATED STATISTICS BEHIND ‘SAFE’ SELF-DRIVING CARS








⚙️ VENTURE FUNDING IS SLOWING DOWN, BUT IT'S MORE COMPLICATED WITH AI





View more

The Deep View

Get our free, 5-minute daily newsletter that makes you smarter about AI. Read by
190,000+ from Google, Meta, Microsoft, a16z and more.

Home

Posts



© 2024 The Deep View.

Privacy Policy

Terms of Use