OpenAI Offering GPT-4o Mini-Cheaper, Stripped Down Version; Meta Has Looked at Investing Billions in Eyewear Giant; Amazon Partners with Better Business Bureau Over Fake Review Brokers; Apple Says It Didn’t Use YouTube Subtitles for its AI

Posted: July 18, 2024 | Author: | Filed under: Uncategorized | Tags: , , , , | Leave a comment

OpenAI has rolled out a stripped-down version of the ChatGPT-4o large language model. Dubbed GPT-4o Mini, it is claimed to have better accuracy than GPT-4 on tasks, and costs substantially less than GPT-3.5 Turbo. Zdnet.com reports that OpenAI is claiming that the new AI model is “the most cost-efficient small model in the market.” It is worth noting that there aren’t any parameters defining large or small models, so this may just be puffing. At any rate, GPT-4o Mini is priced at 15 cents per million input tokens and 60 cents per million output tokens…vastly more affordable than the previous models and 60% cheaper than GPT-3.5 Turbo. The model only offers text and image support right now, with audio and video expected to be added at a later date. Its training data is current through October 2023. 

Meta has looked at investing billions in eyewear giant EssilorLuxottica. According to theverge.com, the blast of cash would be in furtherance of Meta’s partnership with the owner of Ray-Ban and numerous other eyewear brands. The Ray-Ban Meta smart glasses released last year now support multimodal AI to identify what wearers are seeing. They also sold more in a few months than the previous pair did in two years, according to EssilorLuxottica’s CEO.

Amazon is teaming up with the Better Business Bureau to fight fake review brokers…starting off with a lawsuit against a firm called ReviewServiceUSA.com. Geekwire.com says the suit claims that Review Services  allegedly facilitates the selling of fake positive reviews for products on Amazon listings or Better Business Bureau profile pages. Amazon said it blocked more than 250 million suspected fake reviews in 2023.

We had reported yesterday that Apple was among some firms that used its OpenELM model to train AI on YouTube Subtitles. Now, 9to5mac.com reports that Apple says that’s not true. Cupertino claims that the subtitles were only used for research purposes, and that the YouTube subtitles were not used to power Apple Intelligence. Apple has said that Apple Intelligence models were trained “on licensed data, including data selected to enhance specific features, as well as publicly available data collected by our web-crawler.” Apple also stated that it has no plans to build new versions of the OpenELM model. 

I’m Clark Reid and you’re ‘Technified’ for now.

New Anthropic AI Claims to Best ChatGPT; Amazon Drops Plastic Air Pillows for Recycled Paper in Most Packaging; Apple Intelligence Plans for China Unclear; Cleaning Up Cow Burps to Help Slow Global Warming

Posted: June 20, 2024 | Author: | Filed under: Uncategorized | Tags: , , , , | Leave a comment

OpenAI rival Anthropic has released its latest generative AI model named Claude 3.5 Sonnet. Anthropic claims that its model is better than OpenAI’s latest ChatGPT or the models of Google and Microsoft. Techcrunch.com reports that while this appears to be true, Claude 3.5 Sonnet just barely beats OpenAI’s GPT-4o on the benchmark test they ran. Anthropic claims their model better understands nuanced and complex instructions, in addition to concepts like humor. Like other AI, however, it is still notoriously unfunny. It really is a sizable jump from the previous Anthropic model, but techcrunch.com says not as big a leap as OpenAI made going from GPT-3 to GPT-4. Most observers think we will see some large improvements in generative AI yet this year…which will hopefully avoid issues like the fellow at a McDonald’s who ordered some nuggets and the chatbot put in over 200 orders of them! Thankfully, there was human intervention!

In a move that could have a substantial effect on better recycling, Amazon has announced that they have eliminated 95% of the inflated plastic pillows used to cushion all the goodies that are shipped to us. According to geekwire.com, Amazon claims this is “largest plastic packaging reduction effort in North America and will avoid nearly 15 billion plastic air pillows annually.” On a personal note, I’m delighted. It’s not only better to use recycled paper for the environment, but no more deflating those stupid plastic pillows to put them in the garbage. As a bonus, my cat loves the paper almost more than the boxes it comes in. It has been estimated by environmental nonprofit Oceana that Amazon produced 208 million pounds of plastic packaging in 2022…enough to circle the globe over 200 times!

It was a big reveal at WWDC as Apple finally previewed its AI called Apple Intelligence. One thing that is still up in the air though is how they will bring the feature to China. As we reported, Apple will use ChatGPT for their most intense AI queries…and ChatGPT is not approved by the government in China. Now, 9to5mac.com says Apple is talking with Baidu, Alibaba, and another AI group to fill the gap. Another question is how Apple will be able to operate its Private Cloud Compute in China. Samsung has reportedly worked with Baidu, but hasn’t been happy with their AI performance. The net is, Apple Intelligence may debut this fall in the US and elsewhere, but for now…it looks like China will have to wait. 

As has been known for a while, livestock cause about 14.5% of greenhouse gasses…more than all the cars and trucks in the world combined. While it’s funny to say ‘cow burps,’ it remains a serious problem. Cattle provide us with both meat and milk…and that isn’t something most folks are willing to do without…to say nothing about the livelihood of farmers and stockmen who raise the beasts. Well arstechnica.com is reporting a way that could help reduce methane emissions from livestock by over 30%. It’s a synthetic feed additive that absorbs some of the hydrogen produced in the rumen of the livestock before methanogens can use it. It is already approved in the US and Canada. Other additives like red algae are also being looked at…that one can reduce emissions by up to 90%..BUT…it is toxic, and it’s not clear that alterations to get rid of the toxicity are safe in the long term. Still, the approved additive knocking down gasses by 30% is a really good start. It all makes me feel less guilty about grilling those steaks on the 4th of July!

I’m Clark Reid and you’re ‘Technified’ for now.

Apple & Open AI-No Payments For ChatGPT on Apple; YouTube Further Tightens Screws on Ad Blocking; UK Startup Has Rare-Earth Free Magnet for EVs; WhatsApp Will Support 32 Person Video Calls

Posted: June 13, 2024 | Author: | Filed under: Uncategorized | Tags: , , , , | Leave a comment

It has come out in legal filings that Apple has been collecting a stupendous amount of cash from Google-reportedly about $20 billion- to keep Google’s search as the default on Apple products. There has been speculation about how much Apple might be paying OpenAI to run ChatGPT on the latest iPhones and Macs. It turns out that Apple is paying zero. Appleinsider.com reports that OpenAI is doing it for the exposure to the billions of Apple users…for now. The reverse is true also…OpenAI isn’t paying Apple a dime, either. This deal could change in the future, and in fact Apple is still talking with Google about using their AI at some point. Apple Intelligence, the Apple branding for their AI, will run on device with iPhone 15 Pro and Pro Max, and on Macs running M1 chips or better. For more complex queries, limited info will go to Apple Cloud Compute…Apple’s super private and secure system. Queries that exceed what can be done there will then go to ChatGPT…but only after being authorized by the user. OpenAI claims it isn’t collecting any data from Apple Users. 

YouTube has been relentless in their efforts to make it harder and harder to block ads. According to androidpolice.com, the latest move is a move by YouTube to server-side ad injection. This complicates ad delivery and delivery speed, but will make it tougher for now for ad blocker makers to help users skip ads. For YouTube, it also complicates things, as ad info will have to be sent to Premium members, so their client app can skip the ads. Expect ad blockers to figure out a way to take advantage of this to try again to block ads for non-premium members. This is not unlike the almost perpetual race between speed radar makers and radar detector makers that have played cat and mouse for decades. If the new system…being trialed right now… proves to cause too much lag time or hassle, YouTube may lose premium subscribers over it. 

A startup in the United Kingdom has used AI to uncover a new way of making rare earth-free magnets for EVs. Thenextweb.com says Materials Nexus out of London had an algorithm analyze over 100 million combinations of materials to come up with a viable rare earth-free magnet. The reason this is a biggie is that we will eventually run out of rare earths like dysprosium and neodymium. also, a lot of them are mined in China, which makes the supply insecure should the Chinese decide to cut off exports. The substances are crucial for the magnets in the electric motors that power electric vehicles, among other uses…including microchips and superconductors. 

WhatsApp is rolling out an update that…among other things…will allow up to 32 people on a video call. TechCrunch.com reports that they are also adding to the screen sharing they introduced last year with shared audio also available. Previously, you could have 32 on mobile, but now you can have 32 participants on Windows and Macs, too…can you say getting into Zoom’s britches a bit more? Meta has also introduced Meta Low Bitrate codec for WhatsApp to improve call reliability where a user has lousy network connectivity or is using an old device. 

I’m Clark Reid and you’re ‘Technified’ for now. 

OpenAI Board Learned of ChatGPT from Twitter; Apple Looks to Black Box AI Cloud Data for Users; Founder Calls for ‘Memetech’ 

Posted: May 29, 2024 | Author: | Filed under: Uncategorized | Tags: , , , , | Leave a comment

I had originally done a story about a company starting with G, and how there was a large leak about their search algorithm, but after an hour of checking, they hadn’t ok’ed the video…a subtile way of censoring it looks like. Anyway, you will have to check elsewhere for that info. Sorry.

After all the drama of the OpenAI board canning CEO Sam Altman, then hiring him back, more trickles out about what was going on. With ChatGPT being hyped as the biggest thing since fire or electricity, it turns out OpenAI was kind of a snake pit. According to arstechnica.com, in an interview with ‘The Ted AI Show,’ former OpenAI board member Helen Toner said that the OpenAI board was unaware of the existence of ChatGPT until they saw it on Twitter. She went on to share that many at OpenAI were afraid to cross Sam…partially in fear that the company would fall apart. Toner’s main argument is that OpenAI hasn’t been able to police itself despite claims to the contrary. “The OpenAI saga shows that trying to do good and regulating yourself isn’t enough,” she said. This doesn’t bode well for the so-called ’Safety Committee’ formed at OpenAI being able to police anything. 

Although rumors have Apple doing a good deal of their AI right on the latest iPhones in order to keep your data secure, it is expected that at least some information crunching will be done in the cloud. The Information says that Apple plans to process AI applications in a virtual ‘black box’ in the cloud…one that even their employees will be unable to access. This is probably pretty accurate info, as it comes from 4 former Apple employees who worked on the project of integrating AI into Siri. Apple has been working secretly on AI for 3 years, but are still considered to be playing catch up with Google, OpenAI, and Microsoft.

Oh, here’s what we need….better and easier to modify and send memes. As a heavy user of memes, even I have to be skeptical. That said, Alex Taub, a founder of a number of tech startups says it’s time to disrupt the meme ‘industry.’ Techcrunch.com reports that Taub notes that memes are a key component of our online communications. He also says of course it isn’t a necessary thing….but goes on to say that neither are smart refrigerators! Taub says most people that use memes have a meme folder…yep, have a huge one myself. He forsees a program to catalog your memes so you can pull and send an appropriate one to a friend or post it in a good moment. Not just one to talk, Taub has launched Meme Depot…which will be a comprehensive archive of any meme imaginable! Right now he is funding Meme Depot from his past ventures, but expects if it takes off he will make some money from it either by subscription or ads…just like most of the rest of the internet sites do.

I’m Clark Reid and you’re ‘Technified’ for now. 

Google I/O Recap; Open AI’s New GPT-4o; Feds Investigate Waymo Driverless Cars; Apple & Google Bow Cross-Platform Anti-Tracking

Posted: May 14, 2024 | Author: | Filed under: Uncategorized | Tags: , , , , | Leave a comment

Google made a flood of announcements today during the 2 hour I/O Keynote. Of course, it was all heavy on AI. 9to5google.com reports that Google Lens will now get the ability to search with a video. You can shoot a video, ask a question about something in it, and the AI will try to find appropriate answers on the web and serve them. The new Google AI model is Gemini 1.5 Flash. That’s supposed to be a reference to its quickness, not to the old Adobe Flash Player, or the comic book character, or someone running naked across a stage…although that might have gotten some shock value into the presentation today! Anyway, Flash is multimodal, and just as powerful as Gemini 1.5 Pro according to Google. They have also doubled 1.5 Pro’s context window to 2 million tokens. Gemini is being rolled out to the sidebar in Docs, Sheets, slides, Drive, and Gmail when it gets to paid subscribers next month. They claim it will be a general purpose assistant in Workspace that will fetch info from your drive, help you wrote, or give you reminders.

Google also touted Project Astra is a multimodal AI assistant that the company hopes will become a do-everything virtual assistant that can watch and understand what it sees through your device’s camera, remember where your things are, and do things for you. The Google answer to OpenAI’s Sora is a new generative AI model that can output 1080p video based on text, image, and video based prompts. Google is also bowing a custom chatbot creator called Gems that you can customize. Circle to search now can help solve math problems…it won’t do it for you, (so school kids can’t use it to cheat) but will break down problems into easier steps. Something that will affect everyone is AI Overviews…formerly the ‘Search Generative Experience.’ Yep, Google is dropping more AI into their bread and butter search engine. 

Yesterday, getting the drop on Google somewhat, OpenAI released GPT-4o, a new flagship AI model. According to techcrunch.com, it is a rolling release and will hit developer and consumer facing products over the next few weeks. What is it? Well, according to OpenAI, it provides GPT-4 level intelligence but improves on GPT-4’s capabilities across text and vision as well as audio. OpenAI stressed the importance of voice and vision as the large language model interacts more with people…so be sure to say and think nice things about our coming AI and robot overlords. One interesting wrinkle…you can interrupt it as it is giving you an answer, and ask more or clarify, and the chatbot will theoretically be able to handle that. 

The National Highway Traffic Safety Administration has been hot on the case of GM’s Cruise vehicles, which had to stop operation in San Francisco after a series of accidents. Now, the feds are looking into ‘unexpected behavior’ by Waymo self-driving cars. Arstechnica.com says that some of the incidents were reported to the government by Waymo, and others came from the public. The feds are looking into what they call  single-party crashes into “stationary and semi-stationary objects such as gates and chains” as well as instances in which Waymo cars “appeared to disobey traffic safety control devices.” This initial probe is the first step before the NHTSA can issue a potential recall. Earlier this year Waymo voluntarily recalled some 400 self drivers after back to back crashes in Arizona. 

As has been promised since last year, Apple and Google are finally rolling out cross-platform anti-tracking ability. Apple has had this feature for a couple years…it aims to prevent someone using one of their Air Tags to track or stalk someone else. Engadget.com notes that Apple and Google have been collaborating to make it possible to spot and end this kind of behavior across Apple and Android devices, to protect users from unwanted Bluetooth trackers snooping around on them. When an unknown Bluetooth device is seen moving with someone over a period of time, they’ll get an alert that reads “[Item] Found Moving With You,” no matter which platform the tracker is paired with. Apple and Google are rolling out the capability in iOS 17.5 and across Android 6.0 and later devices starting today.

I’m Clark Reid and you’re ‘Technified’ for now.

Google Prohibiting Sites & Apps that Generate Deepfake Porn; ChatGPT Search Engine Rumored Imminent; Threads Now Lets You Control Who Can Quote Your Post; Jack Dorsey No Longer on Bluesky Board

Posted: May 6, 2024 | Author: | Filed under: Uncategorized | Tags: , , , , | Leave a comment

Google is going to prohibit ads promoting websites and apps that generate deepfake porn…starting May 30th. Engadget.com reports that Google has already had strong restrictions in place for ads that feature certain types of sexual content, this update leaves no doubt that promoting “synthetic content that has been altered or generated to be sexually explicit or contain nudity” is in violation of its rules. Advertisers who violate the rules will be suspended without warning. Such deepfake ads are already prohibited in Shopping ads. Hopefully, Google will really police this, and in a manner where it doesn’t end up like whack a mole. 

When it comes to the internet and tech companies, people are always scouring every single word and character to try to determine what’s coming…whether a feature or ‘the next big thing.’ Now, according to mashable.com, a post in Y Combinator’s Hacker News community noted a domain name and security certificate for ‘search.chatgpt.com’ has been created. This could mean we are going to see a ChatGPT search engine sooner not later. Google, the 800 pound gorilla of search already is powered by an AI algorithm…but a ChatGPT one from OpenAI could really juice up the competition. It’s conceivable that before long, one could do a Google search, a ChatGPT search, and one with Microsoft Copilot…and I wouldn’t put it past some brilliant hackers out there to come up with a way to synthesize then streamline the results from all three…for a supercharged search like we’ve never experienced!

Threads is giving more control over who can quote their posts. Engadget.com says if you want to limit quoting your posts, you can limit it to only people you follow…or you can set it so no one can quote your posts at all. The update was announced over the weekend, and Threads is doing it to ‘help keep Threads a more positive place.’ Threads has now climbed past 150 million monthly users. Another recent feature lets users archive posts…either manually or automatically after a selected expiration date.

Jack Dorsey Tweeted over the weekend (not saying post on X, since it was Twitter when Jack ran it) that he is no longer on Bluesky’s board. TechCrunch.com notes that in fact, Jack was the platform’s most prominent backer…he first announced it back in 2019 when he was still CEO at Twitter. Dorsey didn’t elaborate, just replying to a question about him being on the board with a curt “No.” 

I’m Clark Reid and you’re ‘Technified’ for now. 

FCC Fines AT&T, Verizon, Sprint & T-Mobile-Sharing Location Data; Meta is Offering Some Creators Thousands; Eight News Publishers Sue OpenAI & Microsoft-Copyright Infringement; DARPA’s Wild Self-Driving Robot Tank

Posted: April 30, 2024 | Author: | Filed under: Uncategorized | Tags: , , , , | Leave a comment

The FCC has whacked the big mobile carriers in the US a combined $200 million for what the Commission says was illegally sharing customer location data without consent. Theverge.com reports that the carriers “sold access to its customers’ location information to ‘aggregators,’ who then resold access to such information to third-party location-based service providers.” The agency says the carriers effectively “attempted to offload” their responsibility to get customers’ consent to share their location data with “downstream recipients.” Even after being made aware of the issue, the FCC claims, the carriers still failed to limit access to the information. T-Mobile got whacked the hardest…with an $80 million fine. AT&T was number two, getting dinged for $57 million, and Verizon was hit for $47 million. Sprint, which merged into T-Mobile after the investigation started, owes $12 million. All the carriers are expected to appeal. 

Meta is waving thousands of dollars under the noses of some creators if they go viral on Threads. According to engadget.com, the money is part of a new invitation only bonus program. It is “based on the performance of your Threads posts” or “the number of posts you create.” Meta hasn’t given details about how the bonus program is structured, but so far, it appears that the bonuses are tailored to each individual creator. Meta says it is just testing the payment feature, so we don’t know how much it might be expanded…but a couple of creators have gotten offers of “up to $5,000” for Threads posts or replies with 10,000 views or more. That’s not nearly as high as the $10,000 bonuses Reels creators could once earn on Instagram, but not too shabby, either. 

Eight US news publishers have sued Microsoft and openAI, making the claim that the companies are using their copyrighted articles to train generative AI like the ChatGPT series and what Microsoft has recently dubbed Copilot. Cnbc.com says the suit claims the chatbots have  been “purloining millions of the publishers’ copyrighted articles without permission and without payment.” Newspapers operated by the companies that have sued include New York Daily News, the Chicago Tribune, the Orlando Sentinel, the Sun Sentinel in Florida, The Mercury News in California, The Denver Post, The Orange County Register in California and the Pioneer Press of Minnesota. The complaint filed states that “The current GPT-4 LLM will output near-verbatim copies of significant portions of the publishers’ works when prompted to do so.” Previously, the New York Times had sued OpenAI over ChatGPT using its copyrighted property without payment. OpenAI’s CEO Sam Altman said at the time that the suit was without merit, and that the startup had wanted to pay the Times.  

The folks at DARPA, who years ago developed DARPANet…that became the internet…have a wild new project. Bgr.com reports they are testing a self-driving robot tank. The prototype doesn’t have cannons on it, but it is freaky seeing the thing bound over all kinds of difficult terrain at 25 mph with no human input. It has two large, green lighted slits at the front that are status indicators, but look like eyes of some green monster. The tank is part of DARPA’s so-called RACER fleet….based on their RACER heavy platform that can be used for tanks, personnel carriers and other vehicles, including the Manta Ray underwater drone which bowed recently. 

I’m Clark Reid and you’re ‘Technified’ for now. 

ChatGPT-No Account Required; Yahoo Picking Up AI Powered News App; Microsoft Working on Xbox AI Chatbot; Google Must Destroy Browsing Data Collected-Settlement

Posted: April 2, 2024 | Author: | Filed under: Uncategorized | Tags: , , , , | Leave a comment

In a somewhat surprising move, OpenAI has announced that users will no longer be required to create an account to use ChatGPT, its AI chatbot. Bgr.com reports that the move was announced in a blog post. The company said they will be rolling out the change gradually, but didn’t elaborate on a timeline for when it will be widely accessible without an account. Right now, Open AO says over 100 million people a week in 185 countries use ChatGPT. They aim to expand the number of users dramatically by dropping the account barrier. Keep in mind that you will still need a paid subscription to use the latest and greatest GPT 4, and to save and review your chat history, share chats with others, and use voice chat. This may be OpenAI’s version of a loss leader at the grocery store!

Yahoo is in the process of acquiring Artifact, and AI powered news app started by co-founders of Instagram Mike Krieger and Kevin Systrom. According to TechCrunch.com, Systrom and Krieger will work with Yahoo in an advisory capacity through the transition, but Artifact will stop operating as a stand alone app…its tech will be integrated into Yahoo and the Yahoo News app in the next few months. Artifact had said it was starting winding down operations as the market wasn’t big enough to continue further investment. Financial details have not been disclosed. 

What would you think of an AI powered Clippy annoyingly trying to assist you on your Xbox? Ok, I made that part up, but Microsoft is actually testing a new AI powered chatbot for Xbox that can be used to automate support tasks. Theverge.com says that the AI uses an ‘embodied AI character’ which animates when responding to Xbox support queries. The chatbot is plugged in to Microsoft’s support documents for the Xbox network and system. The chatbot can reply to questions and even process game refunds from Microsoft’s support site. Wait ‘till people try to argue with the chatbot about a refund or try to escalate to talk to a manager! 

Under the settlement of a class action suit from 2020 brought by Google Chrome Incognito users, Google will have to destroy ‘billions of data points’ that it improperly collected. Engadget.com reports that the search giant will also have to update data disclosures and maintain a setting that blocks Chrome’s third party cookies by default for the next 5 years. The suit claimed Google told users their info was private in Incognito mode…all the while it was monitoring their activity. Google had argued that Incognito doesn’t mean ‘invisible,’ and that sites could still see their activity. 

I’m Clark Reid and you’re ‘Technified’ for now.

Microsoft Separates Teams from Office Globally; AT&T-Huge Data Breach; OpenAI Teases Synthetic Voice Engine; Apple Says Latest AI Model ReALM is Better than GPT4

Posted: April 1, 2024 | Author: | Filed under: Uncategorized | Tags: , , , , | Leave a comment

Microsoft is going to sell its Teams chat and video app separately from Office worldwide now. Reuters.com reports that the announcement comes 6 months after Redmond was forced to unbundle Teams from Office i Europe to avoid an EU antitrust fine. The EU has been investigating Microsoft’s tying of Teams and Office since 2020. Teams has been a part of Office 365 since 2017 at no extra charge. Rivals have complained that the bundling has given Microsoft an unfair advantage. Starting April 1, customers can either continue with their current licensing deal, renew, update or switch to the new offers. For new commercial customers, prices for Office without Teams range from $7.75 to $54.75 depending on the product while Teams Standalone will cost $5.25. The figures may vary by country and currency. 

In the event you hadn’t heard, AT&T has reset millions of passwords for accounts after a huge data breach from 2021 was dumped onto the dark web in March. According to techcrunch.com, some 7.6 million out of the around 73 million accounts were affected. The larger number includes some former AT&T account holders. AT&T customer account passcodes are typically four-digit numbers that are used as an additional layer of security when accessing a customer’s account, such as calling AT&T customer service, in retail stores, and online. The leaked data includes AT&T customer names, home addresses, phone numbers, dates of birth and Social Security numbers. Of course, AT&T put out the usual corporate bs about how they take data safety seriously. Uh, huh…all you big companies do. 

Open AI launched voice capabilities in Chat GPT last September. Mashable.com notes that now they are previewing their Voice Engine, a model that can take a 15 second audio clip and text prompt and generate longer audio. In other words, voice cloning. Actually, this has been around for a while from other tech firms, and that’s one of the reasons the SAG-AFTRA union demanded and got protection for actors and voice actors to have the actor’s permission and to be compensated when AI clones their voice for other purposes. The big step forward by Open AI is that up to now, it has taken about 15 minutes of recording of a voice to clone it halfway decently. Their new Voce Engine does it with 15 second clip. Right now, the Voice Engine is limited to a very small universe of users. It isn’t just actors and voice actors who are concerned about this tech…recall the fake Joe Biden robocall a couple months ago. The Biden call is thought to hav been made with software from ElevenLabs, not anything from OpenAI. OpenAI claims to be building in safeguards, but the Biden Administration, members of Congress, and other politicians are working to codify safeguards…best not to leave this sort of thing to self-policing. 

We can expect a continuous flow of peeing patch announcements from the tech world, saying basically ‘mine’s bigger than yours, and I can pee further.’ Now, even famously secretive Apple has joined in. Bar.com reports that Apple researchers have published a paper claiming their ReALM large language model is better than ChatGPT4. Apple says its AI can understand and handle contexts of different kinds. Apparently, users can ask about something on the screen of the PC or run in the background, and the large language model can still understand the context and give the correct answer. An example they gave was doing a search for pharmacies. After the list is on screen, the user could ask ‘Call the one on Rainbow Rd,’ or ‘Call the bottom one,’ and the ReALM model could understand and respond. Instead of responding to a text prompt alone like Chat GPT4, the Apple model can use a screenshot and respond from what’s on it. In another article today, some scientists have predicted that ultimately combined AI may end up being like Star Trek’s Borg…only friendlier, they say. While you assimilate that little tidbit….

I’m Clark Reid and you’re ‘Technified’ for now.

Microsoft & OpenAI-Hackers Now Using ChatGPT; Waymo Updates Robotaxi Software After Crashes; Your AI ‘Girlfriend’ = Data-Harvesting Horror Show; Sarah Silverman’s Copyright Suit vs OpenAI Advances

Posted: February 14, 2024 | Author: | Filed under: Uncategorized | Tags: , , , , | Leave a comment

It’s both unsurprising and scary. Microsoft and OpenAI say hackers are already using ChatGPT to improve their cyberattacks. Theverge.com reports that the companies have picked up attempts by Russian, North Korean, Iranian, and Chinese-backed groups using tools like ChatGPT for research into targets, to improve scripts, and to help build social engineering techniques. Hackers are using large language models  to help with “basic scripting tasks, including file manipulation, data selection, regular expressions, and multiprocessing, to potentially automate or optimize technical operations,” according to Microsoft. Both Microsoft and OpenAI say they haven’t detected any ‘significant attacks’ so far. Microsoft is building a Security Copilot, a new AI assistant that’s designed for cybersecurity professionals to identify breaches and better understand the huge amount of signals and data that’s generated through cybersecurity tools daily.

Waymo has voluntarily recalled the software that powers its robotaxi fleet after two vehicles crashed into the same towed pickup truck in Phoenix, Arizona, in December. It’s the company’s first recall. According to techcrunch.com, Waymo said the crashes were ‘minor’ and that neither vehicle was carrying passengers at the time. There were no injuries. Waymo operates its ride-hailing service in Phoenix, San Francisco, Los Angeles, and Austin. All Waymo robotaxis had been updated by January 12th. This has become much more important after competitor Cruise had to suspend operations after a couple accidents…including one in San Francisco where a Cruise car dragged a pedestrian it had hit. 

So, you are involved with an AI romance chatbot, huh? Well, you may not love this. Mozilla checked out 11 different AI romance chatbots, and they all got a ‘Privacy Not Included’ label. Gizmodo.com notes that the chatbots included popular apps such as Replika, Chai, Romantic AI, EVA AI Chat Bot & Soulmate, and CrushOn.AI. Ten of the 11 are selling or can sell your data, too! Some like CrushOn.AI collect info like sexual health, use of medication, and gender-affirming care. 90% of the apps may sell or share user data for targeted ads and other purposes, and more than half won’t let you delete the data they collect. Security was also a problem. Only one app, Genesia AI Friend & Partner, met Mozilla’s minimum security standards. Give yourself some self-love this Valentine’s Day and stay away from these data gobbling apps. 

Although stripped of a couple parts of the complaint, Sarah Silverman’s suit against OpenAI over their training AI models on her books without consent is moving forward. Engadget.com reports that the case’s primary claim that OpenAI directly infringed on copyrighted material by training LLMs on millions of books without permission survives. The court tossed causes of action for negligence, unjust enrichment, DMCA violations, and accusations of vicarious infringement. Other groups suing OpenAI for alleged copyright-related violations include The New York Times, a collection of nonfiction authors (a group that grew after the initial lawsuit) and The Author’s Guild. 

I’m Clark Reid and you’re ‘Technified’ for now. 

← Older Entries
Newer Entries →