Ever since the ChatGPT API opened up, all sorts of apps have been strapping on AI functionality. I've personally noticed this a lot in email clients: Apps like Spark and Canary are prominently bragging about their built-in AI functionality.

The most common features will write replies for you, or even generate an entire email using only a prompt. Some will summarize a long email in your inbox or even a thread. It's a great idea in the abstract, but I think integrations like these conspire to make communication less efficient instead of more efficient. You should feel free to try such featurestheyre fun!but dont expect them to change your life. Here's why.

The Ouroboros of Communication

We are all overwhelmed with email and communication in general. It's easy to look at this as a tech problem because it's happening on screens. It's not a tech problem, thoughat least, it's not only a tech problem. It's a social problem.

You could say that you get too many emails, and that might be accurate. Another way of saying the same thing is that more people are trying to contact you than you feel mentally capable of responding to. Trying to solve a social problem with tech often only creates new social problems.

For example, instead of writing an email myself inviting you to come over and have some beers, suppose I asked ChatGPT to write that email. The result is 220 words long, including an introduction (I hope this email finds you well!), an explanation of the reasons people might want to have beers together (It's the perfect opportunity to catch up, share stories, and simply have a good time), and a few oddly-worded details made up out of thin air (I'll make sure to create a comfortable and welcoming atmosphere, complete with some snacks to complement our beer tasting experience.)

Most people, seeing an email this long, are going to feel too overwhelmed to read it. Maybe they'll use AI on their end to summarize the message. I asked ChatGPT to summarize the long email into a single sentence, and it essentially gave me back my initial prompt: Would you like to come over for beers?

The American philosopher Homer Simpson once called alcohol the cause of, and solution to, all life's problems. AI, in this context, serves a similar function: It creates a problem (the emails are too long) and then solves them (summarizing the emails). It's an ouroboros, a snake eating its own tail, a technology that exists in part to solve the problems it is creating.

It's better, in my opinion, to look at the cultural assumptions instead of reaching for unnecessarily complicated technological ones. What cultural forces are making me think I can't just write a one-sentence email? Can I ignore that, if it makes communication better?

I asked ChatGPT to summarize the long email into a single sentence, and it essentially gave me back my initial prompt: Would you like to come over for beers?

Cultural problems, of course, are harder to grasp than technological ones. You could start sending one-sentence emails right now, but some people might interpret that as rude, or at the very least odd. But any individualor organizationlooking to become more efficient should think about these things. Unless, of course, you want a bot pretending to know that you have beers ranging from local brews to classic favorites in your fridge right now.

We Dont Know The Contexts in Which AI Will Work Best

My friend Kay-Kay and I, for months, had an in-joke that became a ritual: tapping LinkedIn's conversational auto-recommendations. This social network, for some reason, offers suggested replies to messages. It was never not hilarious.

Courtesy of Justin Pot

Read the original:

Apps Are Rushing to Add AI. Is Any of It Useful? - WIRED

Image Credits: nadia_bormotova / Getty Images

Welcome to Startups Weekly. Sign up here to get it in your inbox every Friday morning.

State-sponsored bad actors have long been able to make deepfake videos that are good enough to trick unsophisticated viewers and probably some more clued-in folks, too. That sort of work takes significant processing power and technical know-how to pull off. Now AI is stepping in, handing over an unlabeled glass bottle, muttering, Hold my beer, and cracking its proverbial knuckles. Things that we could barely dream of at the beginning of 2023 are beginning to be possible when it comes to generated video AI.

Of course, with great power comes great responsibility, but tell that to the memelords doin it for the lolz. Personally (and, perhaps, perversely), I think its a great thing that these technologies are making their way into everyones hands. Special effects have been a strange, mythical other that Hollywood does. Generated AI selfies were a rage for a hot minute (is anyone still using Lensa?) and did wonders in educating people on what is possible. Its not that Im excited about this tech being universally available, but (contrary to what these curmudgeonly pieces would indicate), Im an optimist at heart. Perhaps exposing people to whats possible will help give even non-tech-savvy folks a fighting chance at spotting fake videos.

I suppose it is only optimism if its from the optimisme region of France. Maybe what Im experiencing is lightly sparking hope.

Over the past few weeks, Ive done a lot of writing about fundraising for startup founders. In a conversation with a VC this week, I told them that I had a flags-based checklist for evaluating pitch decks (e.g., red flag means that you havent a whelks chance in a supernova to raise funding. I havent figured out if there should be a mauve flag, nor what it would mean). It inspired me to share where founders go wrong when fundraising (TC+). Yes, it means Im showing the world everything I care about in a pitch deck, but, I mean, 100+ articles about pitching and fundraising later, I think that cat was well and truly out of the bag anyway.

You know what early-stage founders really hate? Putting together their traction slide. What do you put when youre straddling that pre-product/pre-revenue line? I had a bit of an epiphany when I was working with one of my pitch clients: Your traction slide, abstractly, is how much risk you have designed out of the business. Tell that story, and you end up with a reasonable traction narrative, even if it isnt directly tied to revenue.

Apropos fundraising, theres been a fair bit of activity on that front:

Its a nice Jobs if you can get it: Apple founder Steve Jobs met his demise from cancer. Now his son, Reed Jobs, takes the wraps off a $200 million venture fund that will back new cancer treatments.

iForgot: Backed by a16z, Rewind launches an iPhone app to help you remember everything.

Dude, wheres my cell tower?: eSIMs are great and all, but you know whats really cool? Being able to pop a local SIM card into your phone and be chillin like a villain, local style. Airalo just raised $60 million to make that a tiny bit easier, even with eSIMs.

This week, the federal government isnt just laying down the law about certain ex-commanders-in-chief. I spent an hour reading the most recent indictment its surprisingly readable, and fascinating AF. The NYT has a great annotated edition. Also, if thats a thing youre interested in, I definitely recommend the Prosecuting Donald Trump podcast. Two extremely experienced lawyers talk about the cavalcade of cluster-copulation thats happening in the legal system. Rather compelling.

Closer to home, in startup land, the Federal Communications Commission (FCC) has been accused of being toothless, but it truly has had enough of one companys BS, fining a robocaller a record $300 million after blocking billions of their scam calls.

Insert inappropriate how long can you go? joke here: It turns out that Tesla has allegedly been a little floppy with the truth about the range estimates for its cars for a hot minute. Suing Tesla is practically a national sport at this time, and, indeed, the first Tesla range inflation lawsuit has been filed.

Lets get to the meat of things: YouTube star MrBeast has a charming, likable, aw-shucks persona, but it turns out he does have finite patience, suing the ghost kitchen behind the MrBeast Burger. Amandas report doesnt include whether you should like and subscribe to the court case.

A HIPPO-sized HIPAA breach: Close to 2.5% of the U.S. population had their health data accessed by MOVEit hackers, a government contractor says.

The social media world continues to be a Muppet wrapped in googly-eye duct tape, or some similarly confusing simile. People truly hate the Twitter-to-X rebrand. How much? Well, Amandas guide for how to make the blue bird come back as your app icon on iOS is right up there with our most-read stories. On top of that, App Store users are decimating Twitters review rating with one-star reviews after the rebrand. Thats ... a lot of steps to not have to stare at an X. Pretty wild: Apple doesnt usually allow one-letter app names, but it made an exception for Tw . . . I mean...X. I avoided throwing myself into the chaos mid-pandemic by deleting the Twitter app off my phone altogether, which is faster and better for your mental health, but Ill leave you to make the best choices for you.

Mammoth > Bird: Famous for its nature programming, U.K. broadcaster BBC is taking a stroll through the digital ecosystem, and it seems it has had enough of Musks shenanigans. Natasha L reports that BBC is testing being on Mastodon, saying that the fediverse is a better fit for public purposes than Twitter or Instagrams Threads.

Robot says youre looking fiiiiine, 0x58 / 0x59: AI really gets its grubby little mitts everywhere, and it seems that Tinder is joining the fray as it tests an AI photo selection feature to help users build profiles. But, as a non-AI, lemme just say: You look great, fam. Id swipe on you. Raaawr.

Uncrop! Enhance: It was a CSI meme, but we are one step closer to being able to uncrop images, revealing whats beyond the edges. Not for real, but based on Photoshops new generative AI feature taking its best guess. And you know what? Its really, really, really good. No wonder every other TikTok video I get served these days seems to feature uncrop shenanigans.

Are you still reading? Your tenacity and persistence are heartwarming. Now, make yourself a cup of tea and pat yourself on the back youve truly mastered the art of the no, this is work, honestly! type of procrastination. I see you. Im proud of you. Youre doing great.

Heres what everyone else has been ogling this past week:

Hacking your way to horsepower: You know what the problem is with selling people $10,000 software upgrades to their cars? At some point, someone is going to change the $GOFAST =0 flag to $GOFAST =1 and get free heated rear seats. Personally, I think its truly ridiculous to turn off the ability to heat rear seats if youve gone through all the trouble of, I dunno, adding the hardware to heat the rear seats, but thats why Im a lowly TechCrunch hack and not the CEO of a car company, a tunnel company, a space company, and whatever X is.

Rolling electric: Fun fact: volvo means to roll. Presumably they mean the wheels and not some sort of sordid MDMA binge, but in any case, the all-electric Volvo EX30 is a huge deal.

Conducting, in your office: What if room-temperature superconductors were real? Tim wondered, and got (1) a really interesting article and (2) a buttload of traffic for his efforts. Well done, Tim. Keep it up, I love reading your stuff.

Get your TechCrunch fix IRL. Join us at Disrupt 2023 in San Francisco this September to immerse yourself in all things startup. From headline interviews to intimate roundtables to a jam-packed startup expo floor, theres something for everyone at Disrupt. Save up to $600 when you buy your pass now through August 11, and save 15% on top of that with promo code STARTUPS. Learn more.

See more here:

AI is going 4-dimensional - TechCrunch

The global COVID-19 pandemic has shown us just how devastating these outbreaks can be and it could have been much worse. Now, scientists have developed an AI application that promises to warn us about dangerous variants in future pandemics.

It's called the early warning anomaly detection (EWAD) system, and when tested against actual data from the spread of SARS-CoV-2, it was accurate in predicting which new variants of concern (VOCs) would emerge as the virus mutated.

Scientists from Scripps Research and Northwestern University in the US used a machine learning method to produce EWAD. In machine learning, vast amounts of training data are analyzed by computers to spot patterns, develop algorithms, and then make predictions about how those patterns may play out in future, unknown scenarios.

In this case, the AI was fed information about the genetic sequences of SARS-CoV-2 variants as infections spread, how frequent those variants were, and the reported global mortality rate from COVID-19. The software could then spot genetic shifts as the virus adapted, usually shown in increasing infection rates and falling mortality rates.

"We could see key gene variants appearing and becoming more prevalent, as the mortality rate also changed, and all this was happening weeks before the VOCs containing these variants were officially designated by the WHO," says William Balch, a microbiologist at Scripps Research.

The specific technique used here by the team is called Gaussian process-based spatial covariance, which essentially crunches the numbers on a set of existing data to predict new data using not just the averages of the data points but also the relationships between them.

By testing their model on something that's already happened and finding close matches between the real and the predicted data, the scientists could prove EWAD's effectiveness at predicting how measures such as vaccines and mask-wearing could cause a virus to continue evolving.

"One of the big lessons of this work is that it is important to take into account not just a few prominent variants, but also the tens of thousands of other undesignated variants, which we call the 'variant dark matter,'" says Balch.

The researchers say their AI algorithms were able to spot "rules" of virus evolution that would otherwise have gone undetected, and that could prove vital in combating future pandemics as they emerge.

Not only that, but the system developed here could also enable scientists to understand more about the very basics of virus biology. That could then be used to improve treatments and other public health measures.

"This system and its underlying technical methods have many possible future applications," says mathemologist Ben Calverley from Scripps Research.

The research has been published in Cell Patterns.

View original post here:

AI May Be Able to Warn Us Before The Next Pandemic Strikes - ScienceAlert

Opinion

Hitching a struggling media industry to the wagon of AI wont serve our interests in the long run

Before we start, I want to let you know that a human wrote this article. The same cant be said for many articles from News Corp, which is reportedly using generative AI to produce 3,000 Australian news stories per week. It isnt alone. Media corporations around the world are increasingly using AI to generate content.

By now, I hope its common knowledge that large language models such as GPT-4 do not produce facts; rather, they predict language. We can think of ChatGPT as an automated mansplaining machine often wrong, but always confident. Even with assurances of human oversight, we should be concerned when material generated this way is repackaged as journalism. Aside from the issues of inaccuracy and misinformation, it also makes for truly awful reading.

Content farms are nothing new; media outlets were publishing trash long before the arrival of ChatGPT. What has changed is the speed, scale and spread of this chaff. For better or worse, News Corp has huge reach across Australia so its use of AI warrants attention. The generation of this material appears to be limited to local service information churned out en masse, such as stories about where to find the cheapest fuel or traffic updates. Yet we shouldnt be too reassured because it does signal where things might be headed.

In January, tech news outlet CNET was caught publishing articles generated by AI that were riddled with errors. Since then, many readers have been bracing themselves for an onslaught of AI generated reporting. Meanwhile, CNET workers and Hollywood writers alike are unionising and striking in protest of (among other things) AI-generated writing, and they are calling for better protections and accountability regarding the use of AI. So, is it time for Australian journalists to join the call for AI regulation?

The use of generative AI is part of a broader shift of mainstream media organisations towards acting like digital platforms that are data-hungry, algorithmically optimised, and desperate to monetise our attention. Media corporations opposition to crucial reforms to the Privacy Act, which would help impede this behaviour and better protect us online, makes this strategy abundantly clear. The longstanding problem of dwindling profits in traditional media in the digital economy has led some outlets to adopt digital platforms surveillance capitalism business model. After all, if you cant beat em, join em. Adding AI generated content into the mix will make things worse, not better.

What happens when the web becomes dominated by so much AI generated content that new models are trained not on human-made material, but on AI outputs? Will we be left with some kind of cursed digital ouroboros eating its own tail?

Its what Jathan Sadowski has dubbed Habsburg AI, referring to an infamously inbred European royal dynasty. Habsburg AI is a system that is so heavily trained on the outputs of other generative AIs that it becomes an inbred mutant, replete with exaggerated, grotesque features.

As it turns out, research suggests that large language models, like the one that powers ChatGPT, quickly collapse when the data they are trained on is created by other AIs instead of original material from humans. Other research found that without fresh data, an autophagous loop is created, doomed to a progressive decline in the quality of content. One researcher said were about to fill the internet with blah. Media organisations using AI to generate a huge amount of content are accelerating the problem. But maybe this is cause for a dark optimism; rampant AI generated content could seed its own destruction.

AI in the media doesnt have to be bad news. There are other AI applications that could benefit the public. For example, it can improve accessibility by helping with tasks such as transcribing audio content, generating image descriptions, or facilitating text-to-speech delivery. These are genuinely exciting applications.

Hitching a struggling media industry to the wagon of generative AI and surveillance capitalism wont serve Australias interests in the long run. People in regional areas deserve better, genuine, local reporting, and Australian journalists deserve protection from the encroachment of AI on their jobs. Australia needs a strong, sustainable and diverse media to hold those in power to account and keep people informed rather than a system that replicates the woes exported from Silicon Valley.

Samantha Floreani is a digital rights activist and writer based in Naarm

{{topLeft}}

{{bottomLeft}}

{{topRight}}

{{bottomRight}}

{{.}}

See the rest here:

Is artificial intelligence a threat to journalism or will the technology destroy itself? - The Guardian