Tech Brew Ride Home - Tue. 01/30 – The Vision Pro Reviews Are Here
Episode Date: January 30, 2024Elon Musk says Neuralink implanted its device into a human for the first time. You can search for Taylor Swift X again, but was a Microsoft tool to blame for the deepfakes? Better AI coding from Meta.... And I read all the Apple Vision Pro reviews so you don’t have to. Sponsors: Notion.com/ride Links: Elon Musk’s Neuralink implants brain tech in human patient for the first time (CNBC) Microsoft Closes Loophole That Created AI Porn of Taylor Swift (404 Media) Meta releases ‘Code Llama 70B’, an open-source behemoth to rival private AI development (VentureBeat) Apple Vision Pro review: magic, until it’s not (The Verge) Apple Vision Pro review: A revolution in progress (Tom's Guide) Apple Vision Pro Review: The Best Headset Yet Is Just a Glimpse of the Future (WSJ) Apple’s Vision Pro is nearly here. But what can you do with it? (Washington Post) Apple Vision Pro Review: A Mind-Blowing Look at an Unfinished Future (CNET) YouTube Video Of My Interview With Chris Dixon 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 Tuesday, January 30th, 2024. I'm Brian McCullough today. Elon Musk says Neurrelink implanted its device into a human for the first time. You can search for Taylor Swift on X again, but was a Microsoft tool to blame for the deepfakes? Better AI coding from meta, and I read all the Apple Vision Pro reviews so you don't have to. Here's what you missed today in the world of tech. Elon Musk has revealed that Neurrelink implanted its device in a human for the first time on January 28th. And the
that the first Neurrelink product is going to be called telepathy.
Musk made this announcement on X, saying, quote,
the first human received an implant from at Neurrelink yesterday and is recovering well.
Initial results show promising neuron spike detection, end quote.
Quoting CNBC, the inhuman clinical trial marks just one step on Neurrelink's path toward commercialization.
Medical device companies must go through several rounds of intense data safety collection and testing
before securing final approval from the FDA.
Neurlink did not disclose how many human patients will participate in its initial inhuman trial.
The company did not immediately respond to CNBC's request for comment about the recent procedure.
As part of the emerging brain computer interface or BCI industry, Neurlink is perhaps the best-known
company in the space thanks to the high profile of Musk, who is also the CEO of Tesla and SpaceX.
A BSI is a system that deciphers brain signals and translates them into commands for external technologies
and several companies like Synchron Precision Neuroscience, Paradromics, and BlackRock NeuroTech have also created
systems with these capabilities.
Parodromics is aiming to launch its first trial with human patients in the first half of this year.
Precision Neuroscience carried out its first inhuman clinical study last year, a patient who
received Synchron's BCI used it to post from CEO Tom Oxley's Twitter account back in 2021, end quote.
Neurrelink's implant is designed to enable severely paralyzed individuals to operate external
devices through brain signals alone. In the fall, the company started enrolling participants for
its inaugural human trial following the FDA's authorization in May. This innovation, if successful,
could enable individuals with severe degenerative conditions such as ALS to interact with communication
tools and social media platforms, controlling cursors, and typing using only their thoughts.
Or as Musk put it on X, quote, imagine if Stephen Hawking could communicate faster than a speed typist
or auctioneer. That is the goal, end quote.
You can search for Taylor Swift on X again. X has lifted its ban on those search terms,
but, quote, we'll continue to be vigilant following the spread of explicit AI-generated images of the singer.
But what I found interesting is where these images allegedly were coming from.
Microsoft has added safeguards to its AI text-to-image tool designer after a report of 4chan users misusing the tool
and indications that that's where the Swift images were generated.
Quoting 404 Media.
Microsoft made the changes.
after 404 media reported that the AI-generated images of Taylor Swift that went viral on Twitter
last week came from 4chan and a telegram channel where people were using designer to make
AI-generated images of celebrities. We are investigating these reports and are taking appropriate
action to address them. A Microsoft spokesperson told us in an email on Friday, our Code of Conduct
prohibits the use of our tools for the creation of adult or non-consensual intimate content,
and any repeated attempts to produce the content that goes against our policies may result in
loss of access to the service. We have large teams working on the development of guardrails and other
safety systems in line with our responsible AI principles, including content filtering, operational
monitoring, and abuse detection to mitigate misuse of the system and help create a safer environment
for users, end quote. And quoting Venture Beat. Open AI's Dolly 3, which powers the tool and Microsoft
designer itself, also had built-in technical prohibitions against users asking the AI tool to generate
explicit imagery. Nonetheless, some users were clearly able to get around these prohibitions
using prompt engineering techniques leading to the creation and spread of the swift images.
As for what specific updates were made to the designer AI service, 404 reported that it
previously was possible to prompt the service with, quote, slightly misspelling the name of
celebrities and describing images that don't use any sexual terms, but result on sexually suggestive
images, but that these techniques no longer work. Yet the question remains, if motivated users,
users will keep testing and find workarounds to the new restrictions. Swift has also reportedly
considered taking legal action over the images, though if and who would be named in a lawsuit is still
unanswered for now, end quote. Meta has released Code Lama 70B, a new version of its code
generation model featuring improved code corrections, a variant optimized for Python, and more.
Folks are saying this is the first local LLM that can beat GPT4 when it comes to coding.
Quoting Venture Beat. This updated model can write code in various programming languages such as Python,
C++, Java, and PHP from natural language prompts or existing code snippets, and it can do it faster,
better and more accurately than ever before. Code Lama 70B is a state-of-the-art large language model
that has been trained on 500 billion tokens of code and code-related data, making it more capable
and robust than its predecessors. It also has a larger context window of 100,000 tokens,
which enables it to process and generate longer and more complex code.
Code Lama 70B is based on Lama 2, one of the largest LLMs in the world with 175 billion parameters.
Lama 2 is a general-purpose LLM that can generate text in any domain and style from poetry to news articles.
Code A Lama 70B is a specialized version of Lama 2 that has been fine-tuned for code generation using a technique called self-attention,
which allows it to learn the relationships and dependencies between different parts of the code.
One of the highlights of Code Lama 70B is Code Lama 70B Instruct, a variant that has been fine-tuned
for understanding natural language instructions and generating code accordingly.
This variant scored 67.8 on Human Evol, a benchmark dataset of 164 programming problems that
tests the functional correctness and logic of code generation models.
This score surpasses the previous best results of open models such as Code Gen 16B mono,
29.3 and StarCoder 40.1 and is comparable to closed models such as GPD4, 68.2, and Gemini Pro,
69.4. Code Lama 70B instruct can handle a variety of tasks such as sorting, searching, filtering,
and manipulating data, as well as implementing algorithms such as binary search, Fibonacci, and factorial.
Code Lama 70B also includes Code Lama 70B Python, a variant that has been optimized for Python,
one of the most popular and widely used programming languages in the world.
This variant has been trained on an additional 100 billion tokens of Python code, end quote.
Automating software creation and modification has been one of the fastest areas of AI to break through to real-world use,
enhancing efficiency, accessibility, and creativity and software development.
Users can instruct their computers to write code, modify existing code easily, or convert code between languages.
But code generation is challenging due to code's precision and complexity,
necessitating extensive data computational power and advanced intelligence. Since you have to follow
precise rules and syntax for code to work, in a way, the hallucination problems of AI are more obvious,
and the seemingly natural tendency for LLMs to be creative is actually hampered. Obviously,
Code Lama 70B is trying to fix that. The review embargo broke this morning at 9 a.m. Eastern,
and I read and watched all of them so you don't have to. Here is everybody's first take on the Apple Vision Pro,
and instead of focusing on just one review, I'm going to take bits and pieces from all the reviews.
First up, at the time of this recording, Marquez Brownlee, just has an unboxing video up, so check that out if that's your thing.
Running through everybody else, The Verge, the Wall Street Journal, others, the general consensus seems to be this.
Marvelous display, great hand-and-eye tracking, and it works seamlessly inside the ecosystem.
But again, it's pricey, and different folks had different takes on what the use case would be.
Nilai, of course, did the review at the verge.
They gave it a 7 out of 10.
They noted that video pass-through can be blurry, which a lot of people seem to pick up on.
Maybe Apple oversold that, quoting Neelai.
The goal is for the Vision Pro to be a complete device that can sit right alongside the Mac and the iPad
and Apple's ecosystem of devices and let you get real work done.
You can use Excel and WebEx and Slack in the Vision Pro, and you can also sit back and watch
movies and TV shows on a gigantic virtual 4K HDR display, and you can mirror your Mac's display and just
use the Vision Pro to look at a huge monitor floating in virtual space. It sounds amazing, and sometimes
it is, but the Vision Pro also represents a series of really big tradeoffs, tradeoffs that are
impossible to ignore. Some of those tradeoffs are very tangible. Getting all this tech in a headset
means there's a lot of weight on your face, so Apple chose to use an external battery pack connected
by a cable, but there are other more philosophical tradeoffs as well. As I've been using it for the
past few days, I kept coming up with a series of questions, questions about whether the tradeoffs were worth
it. Is using the Vision Pro so good that I'm willing to mess up my hair every time I put it on? Is it so
good that I want to lug it around in its giant carrying case instead of my laptop bag? Is it so
good that I want to look at the world around me through screens instead of with my own eyes?
Basically, I keep asking if I prefer using a computer in there rather than out here. And as
interesting as the Vision Pro is, there's a long way to go before it can beat out here, end quote.
He said the external battery pack was fine because you're mostly using it stationary.
Here's Joanna Stern picking up on that in her review from the Wall Street Journal, quote.
The battery pack didn't bother me much, even if I looked like a high-tech marionette,
but I did have to charge every two to three hours, so most of the time I plug myself into the wall with the five-foot cord, end quote.
As for computing, as for the visual keyboard, as Gruber and Baratunde told us on the pod recently,
it's basically worthless. You'll want your Mac keyboard, quoting Joanna again.
There is a built-in virtual keyboard so you can type in thin air, but it will be.
will drive you mad for anything longer than a short message. And selecting smaller buttons with a
pinch should be a carnival game. I started getting real work done once I paired the Vision Pro with a
Bluetooth keyboard and mouse. I leveled up again by connecting my MacBook Pro, which let me write this
review on a giant virtual monitor hovering over my desk, end quote. What about your eyes being
projected outward when someone walks up to you? Oh, and those personas, if you try to do some face-timing,
here's Chris Velasco in The Washington Post. When you're wearing Vision Pro and you turn to look at an
actual person or if one wanders into your field of view, you'll see them emerge through your app
windows. And at least in my experience, talking to those folks feels pretty normal since you can see
them as you always would. But of course, you'll look a little odd with a computer strapped to your
face, even when its outer screen is displaying a facsimile of your eyes. Visual indicators on the headset's
outer screen make it clear when you're not actively engaged in a conversation, but products like
this are still so niche that talking to someone wearing a headset just feels strange.
FaceTime video calls are weird on the Vision Pro too because at no point does the other person see your real face.
Instead, they see a blinking, winking computer-generated persona of you.
Creating one involves holding the headset in front of your face as you move your head around,
smile, raise your eyebrows, and close your eyes, but the results are sort of a mixed bag.
My persona's facial features were mostly on point, though the Vision Pro made my bald head look a bit like one of the cone heads.
And even good-looking personas like the ones used by Apple spokespeople I chat with are clearly denizens of the uncanny.
Valley, end quote. Here was Joanna Stern's take. You look awful, my sister said. It's like Botox
from hell, remarked the always kind Jason Gay. Frightening, said my dad, end quote. Everybody
seemed to love the actual computing part of it, the throwing windows on all the walls in front of you.
Here's Tom's guide, quote. For starters, it's a multitasking champ, thanks to the built-in M2 chip
that runs the OS. For example, I launched Safari in the center of my space, then I turn to my left
and asked Siri to launch Slack and the app appeared.
Then I turned to the right of Safari and open Apple Music so I can have that playing in the background.
Even better, it's easy to move apps around in your virtual space by selecting the bar beneath the app,
and you can resize them by staring at the bottom right corner and dragging it with your finger.
If you want to close an app, you can either stare at the X beneath the window or simply ask Siri to close all apps.
You can also go back home at any time by long pressing the digital crown.
You don't get multiple desktop views, but you do get a crystal clear 4K display that renders text crisply,
and you can make that canvas positively huge, so it dwarfs most of the best monitors.
Your keyboard and mouse still work as they normally would with no latency.
You can even use your keyboard in Vision Pro apps if you want.
Yes, only Apple could do this, and I could see myself taking the Vision Pro on business trips
and using it back at the hotel, end quote.
Exactly the use case, as you know, I've been saying Apple needs to lean into, maybe.
Tom's Guide gave it a four out of five stars, by the way.
But what about the entertainment stuff?
CNET says it's the best wearable display they've ever put on, quote.
The 4K resolution per eye micro-Oled display tech Apple uses is basically the retina moment for VR and AR.
It's vivid, richly colored, HDR, and just stunning.
Not only is it good enough for movies, something Apple is touting constantly,
but it's better than any TV in my house.
There are limits, though.
Apple isn't specifically confirming the field of view, but it feels a bit smaller than the MetaQuest 3,
like seeing an amazing monitor through a scuba mask.
The headset's refresh rate is generally around 90 hertz but can go up to 100 hertz.
Some VR headsets can do 120 hertz, but the Vision Pro looks great for film and video and
seems fine for games so far too.
Right now, the closest thing to a killer app the Vision Pro has is its cinema-level video
playback.
That's a hard sell for a $3,499 headset, but if you go for a demo and see it, you'll probably
be just as stunned as I've been.
Movies can float in your room like hovering TVs or be placed in a virtual cinema mode, end quote.
C-Net concluded by rating it 7.8 out of 10.
Everybody has to mention how expensive this is.
Various people complain about the short battery life, the heaviness of the device.
When you wear it over time, too, here's CNET's take.
It's comfy at first, but after half an hour, the headset feels top-heavy and pushes in on my cheeks a bit.
It's fine for short sessions, though, end quote.
Here was Neelai's take, quote.
noticeable thing about the hardware after a while is that it's just heavy. You're supposed to wear this
thing on your face for long stretches of computer time, and depending on which band and light seal you use,
the headset alone, weighs between 600 and 650 grams. I keep joking that the Vision Pro is an iPad
for your face, but it's heavier than an 11-inch iPad Pro at 470 grams, and pushing close to a 12.9-inch iPad
Pro at 682 grams. So in a very real way, it's an iPad for your face. All of the Vision Pros have
is totally front-loaded too, you're just going to feel it after a while, end quote.
So is there a reason for you to run out and buy this right now? Is there a use case for this for most
people? Not down the road, but today. Here was CNET's conclusion, quote, do I believe in the
destination of this mixed reality future? Yeah, I've been writing about it for 10 years. I can see
it coming. Apple Vision Pro is a moment where the ecosystems are starting to finally arrive.
The hardware is hitting levels of audiovisual quality that are truly remarkable, and input
systems are being reinvented. It's an exciting time, and Vision Pro won't be the only product in this
landscape. However, it'll likely be the most influential since the Oculus Rift. That said, it's clearly
not a device you need to get on board with now. The Mac debuted 40 years ago this month, a coincidence
that Apple seems well aware of. The Mac was the birth of modern computing, but few people
had the first Mac. Will the Vision Pro be the first step toward modern spatial computing in mixed
reality as we know it from now on? Maybe. What really makes Vision Pro seem futuristic isn't the
display or the apps, it's the input, eyes and hands. Other headsets have eye tracking and hand tracking,
but none have the combination working as smoothly, subtly, and intuitively as Vision Pro, end quote.
And here was Neely's conclusion in The Verge, quote, the Vision Pro is an astounding product.
It's the sort of first-generation device only Apple can really make from the incredible display
and pass-through engineering to the use of the whole ecosystem to make it so seamlessly useful
to even getting everyone to pretty much ignore the whole external battery situation.
There's a part of me that says the Vision Pro only exists because Apple is so incredibly capable,
stocked with talent, and loaded with resources that the company simply went out and engineered
the hell out of the hardest problems it could think of in order to find a challenge.
There are a lot of trade-offs, though, big trade-offs, not little ones.
And the biggest trade-off of all is that using the Vision Pro is such a lonely experience,
regardless of the weird ghost eyes on the front.
You're in there.
Having experiences all by yourself that no one else can take part in.
After using the Vision Pro for a while, I've come to agree,
with what Tim Cook has been saying for so long.
Head sets are inherently isolating.
That's fine for traditional VR headsets,
which have basically turned into single-use game consoles
over the past decade,
but it's a lot weirder for a primary computing device, end quote.
So our next bonus episode is going to be with A16Z's Chris Dixon once again
about his new book on Web 3 that comes out today, in fact.
And since I already recorded it, it's on YouTube right now if you want to watch it.
Now, link at the very bottom of today's show notes,
or you can wait until this weekend or next whenever I drop it in the podcast feed.
Talk to you tomorrow.
