Tech Brew Ride Home - Fri. 10/16 – You Can Now Hum To Search Music On Google

Starting point is 00:00:00 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 Friday, October 16th, 2020. I'm Brian McCullough today. Google lets you now whistle or hum to search, which sounds crazy I know, or crazy cool. Are we due for one more Apple event this year? The FCC wants to kill Section 230, but actually can't. Could you do machine learning with practically no data to learn off of? And of course, the weekend long read suggestions. Here's what you miss today in the world of tech. Google's got a new hum, to search feature that it says could let you identify that song that's been stuck in your head for days simply by humming or whistling or singing it into Google search. You can be like

Starting point is 00:01:17 do do do do do do do do do do do do do do do do do do do do do do do do do do do do do do do do do do do do do do do do do do do yeah that's Tom's Diner by Suzanne Vega quoting the verge. The new feature is available today in the Google app on both iOS and Android or in the Google Assistant. Just ask Google Google, what's the song, or tap the newly added search a song button, and then hum your earworm. Google will then show you results based on how likely a match it thinks it is, after which you'll be able to tap results to listen to it, just like you would any other song that you looked up in Google Search. Google says that the feature works by using its machine learning models to, quote, transform the audio into a number-based sequence representing the song's melody, end quote,

Starting point is 00:02:01 which it can then compare to existing songs. The company says that it trains these models on a variety of sources including humans singing, whistling or humming, as well as studio recordings, stripping away things like the instruments and vocal quality to focus just on that numeric sequence. Consequently, the hum to search feature should work whether you're tone deaf or have perfect pitch, end quote. I haven't yet tried this out myself, but if it works even half the time, that would be pretty amazing. And yes, you're welcome for me putting Tom's Diner, the most earwormy song ever made into your head today. And quick bit of trivia. That's also the song that they used to refine the original MP3 codec. I did a whole internet

Starting point is 00:02:44 history episode about that. Look that up. I'm seeing rumors beginning to swirl that we might have yet one more Apple event coming to us this year. Because remember, there was some speculation that Apple would announce the first Apple Silicon Macs this week, and they didn't. Well, maybe hold a place on your calendar for November 17, quoting 9 to 5 Mac. John Prosser claims that Apple will hold its next special event on November 17. According to Prosser, Apple has chosen November 17 to hold its next special event where the company will announce new Mac models, including the first Mac with the Apple silicon chip replacing Intel processors.

Starting point is 00:03:27 Prosser also said that Apple will confirm the event to the public a week before on November 10. The company took a similar approach with its special events in September, and October. Prior to this, a Bloomberg report had already mentioned that Apple plans to launch the new Macs in November, end quote. Not sure how reliable processor has proven to be yet in terms of Apple prognostication, but it is worth noting that we are still also waiting on news of those Apple tags and the over-the-ear AirPods Studio as well. So Apple does still have news potentially to make. y'all, I try hard to leave politics out of this podcast, believe me. That's why I've been staying away

Starting point is 00:04:12 from the whole story about the New York Post and Joe Biden's son and Facebook and Twitter taking steps to limit the reach of that story on their platforms because I find that a mostly political story that's only secondarily a tech story. Plus, I don't know where I fall in terms of the argument around is this Facebook and Twitter crossing some sort of line. On the one hand, we want them to be more editorial, right? And on the other hand, we don't want them to be editorial at all. And on the other hand, we do. And I don't know, it's sort of maddening. But I do find myself agreeing with the people who say, if Twitter and Facebook limit the spread of something, that's not censorship. They're not taking anything down. They're not

Starting point is 00:04:57 preventing something from being published, from getting out there. It already has been published. It is out there. It's more that they're choosing not to amplify something that is already out there. And that's, I think, a fair choice. Like, if I decide not to cover a story on this podcast, that's my choice. If I decide to try to stay away from politics on the show, that's my choice. That's not me censoring something. That's me saying, that's not something I choose to talk about. quoting Parker Thompson on Twitter. I have read the argument a lot in the last day that social media is the public square. No, the internet is the public square and it is infinite.

Starting point is 00:05:38 Twitter is an Arby's on the square and you're inside yelling about their lack of vegan options being literal fascism, end quote. So that's the most I'll say about that, if I can help it. But I did want to talk about this. swirling around all of the things like this in the background is the whole discussion of amending section 230 of the Communications Decency Act, the thing that shields platforms like Facebook and Twitter from liability over content users posts on those platforms. Well, coincidentally, yesterday, SEC Commissioner Ajit Pai suddenly said that the FCC will start a rulemaking action to clarify the meaning of Section 230, and

Starting point is 00:06:22 that social media doesn't have, quote, special immunity denied to other media outlets, end quote. Quoting Bloomberg. The announcement came hours after Senate Republicans demanded the chief executive officers of Facebook and Twitter, explain steps their sites took to limit the distribution of a controversial New York Post article concerning Hunter Biden, the son of Democratic presidential nominee Joe Biden. Twitter and Facebook said they did it because of questions about the article's accuracy and use of hacked material, end quote. Problem is, Pi and the FCC have no authority over Section 230, like none, quoting Mike Maznick and TechDirt. Pye is wrong in almost everything he says in his statement.

Starting point is 00:07:05 The FCC has no jurisdiction over Internet websites. Previous lawsuits have already held that. Furthermore, the FCC has no jurisdiction over Section 230, which was explicitly written to deny the FCC any authority over websites. The FCC has no power to reinterpret the law. The final paragraph is the most ridiculous of all. Pye is correct that social media companies have a First Amendment right to free speech, and Section 230, as was written and properly and regularly interpreted by dozens of court decisions,

Starting point is 00:07:35 none of which the FCC has ever said a word about, helps guarantee that right is not diminished through frivolous, bogus, and misdirected litigation. Also, the claim that the immunity is, quote, denied to other media outlets is straight up wrong. Any outlet is protected from liability for third-party content on their websites. It's why Fox News and Breitbart can have comments on their websites. It's why things like Parlor and Gab can exist. Pye knows this. He's just being disingenuous. In terms of actual impact, all this will serve to do is rile people up,

Starting point is 00:08:09 waste a ton of time, and not actually change anything, because it can't. But it will create a huge mess in the meantime, distracting everybody, and wasting a ton of resources, end quote. I actually want to underline something that Mike said there, because a lot of other people have been making this point as well. If your concern is that platforms have a power to quell speech or take things down or limit reach or, quote, censor things, if that's what you believe is happening, if that's your concern, then the last thing you want to do is Monkey with Section 230. Because if platforms are liable to be sued for everything that billions of people post trillions

Starting point is 00:08:45 of times a day, if they're suddenly opened up to lawsuits at scale, then believe me, they will immediately and very aggressively take a ton of things down. If they didn't, they'd end up risking being sued out of existence. You can be unhappy with editorial decisions made by the platforms, I often am, or you can think that Section 230 is a bad thing, I often do. But I'm wondering if it's possible to do both simultaneously. This is the first of the first of the first of the first I've heard of it, but apparently there might be a radical new technique that would allow artificial intelligence to learn with practically no data required to learn off of. Quoting MIT technology review, machine learning typically requires tons of examples. To get an AI model to

Starting point is 00:09:37 recognize a horse, you need to show it thousands of images of horses. This is what makes the technology computationally expensive and very different from human learning. A child often needs to see just a few examples of an object or even only one before being able to recognize it for life. In fact, children sometimes don't need any examples to identify something. Shown photos of a horse and a rhino and told a unicorn is something in between, they can recognize the mythical creature in a picture book the first time they see it. Now, a new paper from the University of Waterloo in Ontario suggests that AI models should also be able to do this, a process the researchers call less than one shot or LO shot learning. In other words,

Starting point is 00:10:18 an AI model should be able to accurately recognize more objects than the number of examples it was trained on. That could be a big deal for a field that has grown increasingly expensive and inaccessible as the datasets sets use become ever larger. The researchers first demonstrated this idea while experimenting with the popular computer vision data set known as Emnist. Emnest, which contains 60,000 training images of handwritten digits from zero to nine, is often used to test out new ideas in the field, end quote. The piece then goes into great detail describing the techniques used, like using soft labels, the machine learning algorithms known as K-nearest neighbors.

Starting point is 00:10:56 It's way too complex to go into now. You'll have to read the piece. But the gist of it is this. If proven out, this technique, quote, could make AI more accessible to companies and industries that have thus far been hampered by the field's data requirements. It could also improve data privacy, because less information would have to be extracted from individuals to train useful models, end quote. Which is why I'm sharing it here, even though it sounds like a long read, because if this proved out, it would be an evolutionary, revolutionary leap in

Starting point is 00:11:26 machine learning. But did someone say long reads? Because it's time for the weekend long read suggestions. Yes, I was a bit harsh on the idea of LiDAR coming to iPhones this week. as you know every iPhone event we hear about the wonders of AR on iPhones and it never really amounts to much. But still, it is worth noting the first three-dimensional LiDar sensor was only introduced a decade ago and it cost $75,000 and it was the size of, I don't know, a cooler. So it's pretty impressive that in 10 short years, LiDAR tech has gotten small enough and cheap enough that it could come to smartphones. So I was interested to learn what made this possible. It's a technology called VC cell. Ars Technica expounds, quote,

Starting point is 00:12:17 The combination of VC cells and spads enables a dramatic simplification of conventional LiDAR designs. Velodyne's original three-dimensional LiDar mounted 64 individually packaged lasers in a column on a spinning gimbal. Each laser had a matching detector. The complexity of this design and the need to precisely align each laser with its corresponding detector was one reason

Starting point is 00:12:40 Velladine's early LiDAR units were so expensive. More recently, a number of companies have experimented with using small mirrors to steer a laser beam in a scanning pattern. This design requires only a single laser instead of 64, but it still involves at least one moving part. By contrast, Apple, Oster, and Ibeo are designing LiDAR sensors with no moving parts at all. With hundreds or thousands of lasers on a chip, VC cell-based LIDARs can have a dedicated laser for each point in the LiDAR's field of view. And because all these lasers come pre-packaged on one chip, assembly is much simpler than for Velodyne's classic spinning design, end quote. Which is why this might end up being a case of LiDAR tech moving from cars to phones and then eventually back to cars again.

Starting point is 00:13:28 Speaking of cars, Fast Company looks at a different revolution that might be forthcoming. Internal combustion cars are super complex machines, with, again, lots of. Lots of of moving parts, brakes, suspension, cooling systems, gas lines. More than 30,000, crucially, moving parts are required to make a combustion car go. But once you're designing cars run by batteries, your need for moving parts is reduced dramatically. In essence, car design is about to be radically transformed because you could basically just design all cars like skateboards, like just a platform that you can put whatever you want on top of. Quote, Delivery giants like Amazon are interested in these electric skateboards to power delivery fleets,

Starting point is 00:14:12 as it just revealed new electric vans built atop a similar electric platform made by Rivian. For Amazon to have 100,000 EVs on the road by 2030, it needs these vehicles to be simple to repair with interoperable parts. The skateboard design ensures a delivery vehicle is never out of commission for long. But what about the cars for the rest of us? With a bit more imagination, you can picture consumer vehicles becoming far more personalized as dozens of of aftermarket companies build varying cabins for a skateboard base. Even if you don't want to buy such a car yourself, small business owners probably will. Vehicles could mobilize the nature of brick-and-mortar retail and services, much like food trucks shook up the restaurant industry in the mid-aughts. Design Studio

Starting point is 00:14:52 New Deal Design has even suggested that linked together individual vehicle storefronts could amass to something like a mobile city that can cruise like a parade or perhaps a mall on wheels. We're well on way to a world of wildly diverse vehicles where design is limited more by the legalese of our road laws than by the creative decisions made by a few big automakers, end quote. Next, one company that you think would be severely hampered by these COVID times is clear, that company that allows you to zip through security lines at airports by scanning your eye and your fingerprint and all that good stuff. But 1-0 says, COVID has actually accelerated clear's probably inevitable pivot to be.

Starting point is 00:15:35 become the digital identity solution for everything. Quote, at its core, clear monetizes trust. When the company verifies a person's identity, whether that be to enter an airport, a stadium, or buy a beer at a concession stand, Clear is affirming that they are who they say they are. Right now, this verification process means priority access to an airport or stadium security line as a trusted clear member. But in the future, documents and slideshows reviewed by 1-0 suggests Clear plans to be the company that verifies your identity every time you will. have swiped a credit card, shown your ID at a door, or handed over a health insurance card, end quote. And we've been speaking about quantum computing lately. But for years, when people have

Starting point is 00:16:17 talked about quantum computing, they also often mentioned as an aside that, you know, once quantum is mastered, it'll be so powerful that it'll be able to break all the encryption that we know of at the moment in like seconds, which seems like a bad thing, right? I mean, that would fundamentally break, not just the internet, but a lot of, you know, modern banking, all sorts of things. Well, from the Wall Street Journal, let me clue you in on the race to protect everything from being, you know, cracked. Quote, the worst case scenario is quite bad, says Chris Pichert, Associate Professor of Computer Science and Engineering at the University of Michigan, who has been studying cryptography for two decades. That's why Dr. Pichert and hundreds of the world's top cryptographers are involved in a competition to develop new encryption standards for the U.S.

Starting point is 00:17:03 which would guard against both classical and quantum computing cyber attacks. This summer, federal officials announced the 15 algorithms that will be considered for standardization, meaning the winners would become part of the architecture of the internet, protecting people's sensitive data. Next, researchers will spend about a year trying to break them to see which ones hold up and test them to get the best balance of performance and security, end quote. And finally, Corey Doctoro's Little Brother series has been a young adult sci-fi staple for generations of teenage hacktivists, people like Aaron Swartz and Edward Snowden. Well, Dr. O's third book in the series is out, and as Wired says, instead of radicalizing young hackers, with this new story,

Starting point is 00:17:47 Dr. O' seems to want to redeem them, quote. Doctor O says the book is meant to stand alone for new readers, even non-techie civilian observers on the sidelines of the crypto wars, but that it's also meant to speak to the core cypher punk audience of the first two Little Brother books. And that includes the ones who didn't turn out to be the heroes of their own story. Quote, a bunch of people who grew up reading Little Brother, imagining that they would become revolutionaries, woke up one day and realize that they're not revolutionaries, that in fact they're helping to make things worse, that they're part of a system that harms people, says Eva Galperin, a longtime digital activist and head of the Electronic Frontier Foundation's Threat Lab. Galperin serves in part as

Starting point is 00:18:29 the inspiration for Masha's character. Both hackers, fictional and non-fictional, were born in the Soviet Union, but grew up in San Francisco with immigrant parents. But Galprin sees Masha also in her idealistic friends, who went to work for Facebook or Palantir or government agencies, vowing to change them from the inside, but finding themselves changed instead. Quote, this is a book for the people who realize that they've grown up and made a lot of compromises, as Galperin says, and about how you turn back from that, end quote. That's all for today, but not for this week, because we do have a weekend bonus episode coming at you tomorrow, so be on the lookout for that. And yes, to follow up, I indeed got the phone

Starting point is 00:19:19 ordered this morning. No problem. Well, 15 minutes of problems, because the Apple Store app just kept crashing, but eventually I got my order in. So, mission accomplished. Talk to you on Monday. Thank you.

Tech Brew Ride Home - Fri. 10/16 – You Can Now Hum To Search Music On Google

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.