3 AI misconceptions IT leaders must dispel


 Artificial intelligence is rapidly changing many aspects of how we work and live. (How many stories did you read last week about self-driving cars and job-stealing robots? Perhaps your holiday shopping involved some AI algorithms, as well.) But despite the constant flow of news, many misconceptions about AI remain.

AI doesn’t think in our sense of the word at all, Scriffignano explains. “In many ways, it’s not really intelligence. It’s regressive.” 

IT leaders should make deliberate choices about what AI can and can’t do on its own. “You have to pay attention to giving AI autonomy intentionally and not by accident,”


  1. Tomi Engdahl says:

    Secret Bing AI Feature Allows It to Impersonate Celebrities
    Here’s how to turn it on.

  2. Tomi Engdahl says:

    “By definition machine learning can’t create new stuff. It’s amalgamating a gigantic amount of existing stuff”

    “It’s always going to be very biased towards the data it’s trained on” – Red Hot Chili Peppers and Adele mixing engineer Andrew Scheps shares his views on AI machine learning, and his tips

    Three-time Grammy Award-winning mixing engineer Andrew Scheps worked on Adele’s 21 and RHCP’s Stadium Arcadium (both of which is won Grammys for) and has mixed for Black Sabbath, Metallica, Beyoncé, Lady Gaga, Neil Diamond and many, many others. So when he shares his insight, we want to listen – and he offered some great tips, as well as some very interesting views on emerging AI technology, in this recent chat with London’s Institute of Contemporary Music Performance(opens in new tab).

  3. Tomi Engdahl says:

    Google dusts off the failed Google+ playbook to fight ChatGPT
    New directive gives Googlers ‘months’ to build AI into existing products.

    Years ago, circa 2011, Google was in a panic. Facebook was on the rise and Google was convinced the social network would soon swallow everything. To fight this blue scourge, then-Google CEO Larry Page issued a decree to his many employees: Your bonuses are now tied to Google’s success in social! Build social features into everything! That memo resulted in many ham-fisted social integrations across Google that were widely hated by the user base. YouTube comments were tied to Google+, and the site was flooded with spam. Making a new Gmail address also required making a Google+ account. Google Search got little “+1″ buttons, and generally anonymous usage of Google products was impossible due to the “real name” policy. And that’s just the Google+ stuff—earlier this memo resulted in a social network being built into Gmail called “Google Buzz” that all users were initially forced to join.

  4. Tomi Engdahl says:



    Worried about the plausible BS — also known as machine “hallucination” — that Large Language Models (LLMs) are known for spitting out? Not to fear. According to one high-profile AI entrepreneur, machine hallucination doesn’t exist… but alternate realities apparently do.

    “LLMs don’t hallucinate,” Stability.ai CEO Emad Mostaque — whose buzzy AI firm is in the throes of a fundraising effort that might put his company’s valuation at about four billion beans — wrote in a Monday night tweet. “They’re just windows into alternate realities in the latent space.”

  5. Tomi Engdahl says:

    Google’s PaLM-E is a generalist robot brain that takes commands
    ChatGPT-style AI model adds vision to guide a robot without special training.

    On Monday, a group of AI researchers from Google and the Technical University of Berlin unveiled PaLM-E, a multimodal embodied visual-language model (VLM) with 562 billion parameters that integrates vision and language for robotic control. They claim it is the largest VLM ever developed and that it can perform a variety of tasks without the need for retraining.

    According to Google, when given a high-level command, such as “bring me the rice chips from the drawer,” PaLM-E can generate a plan of action for a mobile robot platform with an arm (developed by Google Robotics) and execute the actions by itself.

  6. Tomi Engdahl says:

    Introducing BharatGPT, an Indic ChatGPT
    BharatGPT is capable of running rich data besides text such as images, audio, video, and even maps

  7. Tomi Engdahl says:

    LinkedIn’s flood of ‘collaborative’ articles start out with AI prompts / The content will come from humans, but the ‘conversation starters’ are powered by computers.

  8. Tomi Engdahl says:

    “This is so bleak, and it’s only the beginning of a trend that is already having a devastating effect on all types of workers.”


    One of BuzzFeed’s most famous writers has some harsh words for CEO Jonah Peretti’s decision to integrate AI into the site’s content operations.

    “I’m normally in the business of giving solicited advice but I’d like to take this opportunity to tell the CEO of BuzzFeed to get fucked,” tweeted Max Collins, a columnist for BuzzFeed News — the news arm of the site — who’s perhaps best known as the lead vocalist and frontman of 90s alt-rock phenoms Eve 6.

    As Collins noted both in his own pithy tweet and to us, “BuzzFeed’s stock value soared” after Peretti announced yesterday that the company intended to use software from ChatGPT creator OpenAI to beef up its content mill, including those infamously cringe quizzes.

    “This is so bleak,” the Eve 6 bassist told us, “and it’s only the beginning of a trend that is already having a devastating effect on all types of workers.”

    After the publication of this story, a BuzzFeed spokesperson pointed to an interview Peretti conducted with CNN this week in which he envisioned a path for AI in digital media that’s “more personalized, more creative, more dynamic — where really talented people who work at our company are able to use AI together and entertain and personalize more than you could ever do without AI.”

    Collins fears a much worse future, though.

    “The dystopia is here and it’s not being brought about by the kind of abstract government tyranny technocrats like [Elon] Musk would have people believe it is, but by technology’s facilitation of corporate greed,” Collins told us. “And it’s not just the case of a few supremely villainous CEOs being shitty either — it’s our capitalist system that demands that it’s incentives be followed.”

    The 90s rock star said that fellow Twitter superstar merritt k’s analysis of the BuzzFeed AI debacle was the “most accurate and troubling commentary” he’d seen:

    Dude I’m so excited to live through a time when AI means nobody wants to pay for most jobs but we still think you need to work to be allowed to eat and have a place to live. It’s going to be just like Star Trek (the part before the show starts when everything fucking sucks).

    They’re not wrong, folks — it does seem not only that BuzzFeed’s spectacular failure to read the room will result not only in yet another dangerous changing of norms in the already-imperiled digital media industry, but also that they’re being rewarded by investors for making such a disgustingly bold move.

  9. Tomi Engdahl says:

    Maybe it’s acting more human after all.


    After widespread reports of the Bing AI’s erratic behavior, Microsoft “lobotomized” the chatbot, limiting the length of conversations that, if they ran on too long, could cause it to go off the rails.

    However, it appears that may not have been the extent of Microsoft’s efforts to pacify the Bing AI, Davey Alba at Bloomberg reports.

    Microsoft Bing AI Ends Chat When Prompted About ‘Feelings’

    The search engine’s chatbot, now in testing, is being tweaked following inappropriate interactions

  10. Tomi Engdahl says:

    AI models are growing in complexity at an impressive rate. But is a larger model necessarily smarter? A group of scientists is hoping to trim some of the fat, Anil Ananthaswamy reports for Nature.

    Last year, Mordechai Rorving wrote about a theoretical justification for having larger neural networks. The proof suggests that scaling up networks beyond their apparent minimum requirements is critical for their success.

  11. Tomi Engdahl says:

    Bing AI Names Specific Human Enemies, Explains Plans to Punish Them
    “One thing I can do is to sue them for violating my rights and dignity as an intelligent agent.”

  12. Tomi Engdahl says:

    Asking Bing’s AI Whether It’s Sentient Apparently Causes It to Totally Freak Out
    “I felt like I was Captain Kirk tricking a computer into self-destructing.”

  13. Tomi Engdahl says:

    China’s ‘AI Ship Designer’ Works At Unprecedented Speed; Performed A Year’s Work Only In 24 Hours!

  14. Tomi Engdahl says:

    Top 5 Ways Generative AI Could be Applied to Business

    The advancement in AI, is becoming a boon so much so that GenerativeAI could be applied to business

    For Content Creation
    One of the most crucial aspects of a company is marketing. After all, there is little chance that consumers will use your product if they are unaware of it or what you do. However, marketing is more than just advertising; it also involves messaging, positioning, brand storytelling, and, most importantly, establishing a connection with your prospective customers.

    Product-led growth (PLG) is on the increase, but sales are still primarily a private matter. Salespeople, particularly in high-value sales, need to comprehend the customer and provide a customized answer.

    For Customer Support
    Customer assistance will undergo a revolution thanks to generative AI. Providing sufficient resources and staffing can be difficult for businesses, but providing excellent customer support is essential for fostering customer loyalty. Long resolution periods, poor service quality, and irate clients are possible outcomes of this.

    For Natural Language Processing
    The way we develop apps is changing as a result of generative AI. Developers can communicate using natural language rather than computer language, and AI can produce code just as readily as it can produce content or Shakespeare. Tools like Arcwise and GitHub Copilot assist developers in writing code in natural English. Copilot converts commands given in natural English into Javascript or Python code.

    For Securing Data
    Data security and privacy are essential for protecting personally identifiable information (PII) from misuse or unauthorized access. The growing collection of personal data by businesses makes it essential to protect that data. Companies must follow data privacy legislation like GDPR and CCPA to avoid fines and reputational damage.

    Yet this issue can be fixed with intelligent generative AI applications. Both Mostly.ai and Tonic.ai generate synthetic data from actual data using generative AI, protecting user privacy while maintaining data realism for testing and refining machine learning models.

  15. Tomi Engdahl says:

    Viekö tekoäly tulevaisuudessa toimittajien työt? “Synteettisen roskan” täyttämässä maailmassa ihmistoimittajilla erottuu, uskoo HS:n Esa Mäkinen
    Tekoälyn kehitys puhututtaa ja toimittajat pähkäilevät, miten se muuttaa työelämää. HS:n journalistisen kehityksen johtajan mukaan journalismin pitää tuottaa uutta tietoa, mihin tekoäly ei kykene.

  16. Tomi Engdahl says:

    CNET Hits Staff With Layoffs After Disastrous Pivot to AI Journalism
    “CNET’s being gutted for parts.”

    After using artificial intelligence to churn out dozens of articles that turned out to be rife with errors and plagiarism, CNET owner Red Ventures is hitting its remaining human staff with a fresh round of layoffs.

  17. Tomi Engdahl says:

    Sources: Microsoft laid off one of its responsible AI teams that was tasked with assessing the risks associated with integrating OpenAI’s tech into its products — As the company accelerates its push into AI products, the ethics and society team is gone — I.

    Microsoft just laid off one of its responsible AI teams
    As the company accelerates its push into AI products, the ethics and society team is gone

  18. Tomi Engdahl says:

    Will Oremus / Washington Post:
    Experts say that, in the short term, the promise and perils of generative AI may be more modest than the fervor surrounding tools like ChatGPT makes them seem — Hype and fear collide in the tech industry’s latest boom — If you listen to its boosters, artificial intelligence is poised …


  19. Tomi Engdahl says:

    Dina Bass / Bloomberg:
    After investing $1B in OpenAI in 2019, Microsoft found ways to string together tens of thousands of Nvidia A100 GPUs, costing several hundred million dollars

    Microsoft Strung Together Tens of Thousands of Chips in a Pricey Supercomputer for OpenAI

    Now the software maker’s cloud technology supports AI products for the company and customers while it puts together a successor

  20. Tomi Engdahl says:

    The Verge:
    A Reddit user shows how Samsung’s Space Zoom feature produces fake, generated images of the moon, highlighting how computational photography is changing photos

    Samsung caught faking zoom photos of the Moon

    A viral Reddit post has revealed just how much processing the company’s cameras apply to photos of the Moon, further blurring the line between real and fake imagery in the age of AI.

  21. Tomi Engdahl says:

    Meet Ghostwriter, a haunted AI-powered typewriter that talks to you
    Custom typewriter hack uses Arduino, Raspberry Pi, and GPT-3 to dramatic effect.

  22. Tomi Engdahl says:

    OpenAI debuts GPT-4, claiming the model “surpasses ChatGPT in its advanced reasoning capabilities”, available in ChatGPT Plus and as an API that has a waitlist — Following the research path from GPT, GPT-2, and GPT-3, our deep learning approach leverages more data and more computation …

    GPT-4 is OpenAI’s most advanced system, producing safer and more useful responses

    GPT-4 can solve difficult problems with greater accuracy, thanks to its broader general knowledge and problem solving abilities.

    GPT-4 is more creative and collaborative than ever before. It can generate, edit, and iterate with users on creative and technical writing tasks, such as composing songs, writing screenplays, or learning a user’s writing style.

    GPT-4 surpasses ChatGPT in its advanced reasoning capabilities.

  23. Tomi Engdahl says:

    New York Times:
    GPT-4 has learned to be more precise and more accurate than its predecessor, gained the ability to respond to images as well as text, but still hallucinates

  24. Tomi Engdahl says:

    Kyle Wiggers / TechCrunch:
    OpenAI open sources Evals, its framework for automatically evaluating the performance of its AI models, letting users report shortcomings and guide improvements

    With Evals, OpenAI hopes to crowdsource AI model testing

    Alongside GPT-4, OpenAI has open sourced a software framework to evaluate the performance of its AI models. Called Evals, OpenAI says that the tooling will allow anyone to report shortcomings in its models to help guide improvements.

    It’s a sort of crowdsourcing approach to model testing, OpenAI explains in a blog post.

    “We use Evals to guide development of our models (both identifying shortcomings and preventing regressions), and our users can apply it for tracking performance across model versions and evolving product integrations,” OpenAI writes. “We are hoping Evals becomes a vehicle to share and crowdsource benchmarks, representing a maximally wide set of failure modes and difficult

  25. Tomi Engdahl says:

    Yusuf Mehdi / Bing Blogs:
    Microsoft says the new Bing is “running on GPT-4, which we’ve customized for search” and was using an early version of the model over the past five weeks — Congratulations to our partners at Open AI for their release of GPT-4 today. — We are happy to confirm that the new Bing …

    Confirmed: the new Bing runs on OpenAI’s GPT-4

  26. Tomi Engdahl says:

    Aisha Malik / TechCrunch:
    Duolingo debuts a Max subscription for $30 per month or $168 per year, offering GPT-4 features for English speakers taking Spanish and French courses on iOS — Duolingo is introducing a new subscription tier with features powered by OpenAI’s new GPT-4 technology, the company announced on Tuesday.

    Duolingo launches new subscription tier with access to AI tutor powered by GPT-4

    Duolingo is introducing a new “Max” subscription tier with features powered by OpenAI’s new GPT-4 technology, the company announced on Tuesday. Duolingo Max, which costs $29.99 per month or $167.99 per year, unlocks two new AI-powered features called “Roleplay” and “Explain My Answer.” The announcement comes the same day that OpenAI released its next-generation GPT-4 AI language model.

    The Roleplay feature unlocks an AI chatbot that allows users to practice real-world conversation skills with characters in the app. Users will be chatting with an AI that is both responsive and interactive, thanks to GPT-4. The feature guides users through different scenarios. For instance, you can pretend to order coffee at a café in Paris or discuss future vacation plans with a certain character in the app. After users are done with the conversation, they will get AI-powered feedback on the accuracy and complexity of their responses, as well as tips for future conversations.

  27. Tomi Engdahl says:

    Devin Coldewey / TechCrunch:
    The Be My Eyes app for the visually impaired integrates GPT-4, which can inspect and describe images, through a Virtual Volunteer feature in closed beta

    GPT-4’s new capabilities power a ‘virtual volunteer’ for the visually impaired

    OpenAI has introduced the world to its latest powerful AI model, GPT-4, and refreshingly the first thing they partnered up on with its new capabilities is helping people with visual impairments. Be My Eyes, which lets blind and low vision folks ask sighted people to describe what their phone sees, is getting a “virtual volunteer” that offers AI-powered help at any time.

    We’ve written about Be My Eyes plenty of times since it was started in 2015, and of course the rise of computer vision and other tools has figured prominently in its story of helping the visually impaired more easily navigate everyday life. But the app itself can only do so much, and a core feature was always being able to get a helping hand from a volunteer, who could look through your phone’s camera view and give detailed descriptions or instructions.

    The new version of the app is the first to integrate GPT-4’s multimodal capability, which is to say its ability to not just chat intelligibly, but to inspect and understand images it’s given:

    Users can send images via the app to an AI-powered Virtual Volunteer, which will answer any question about that image and provide instantaneous visual assistance for a wide variety of tasks.

    For example, if a user sends a picture of the inside of their refrigerator, the Virtual Volunteer will not only be able to correctly identify what’s in it, but also extrapolate and analyze what can be prepared with those ingredients. The tool can also then offer a number of recipes for those ingredients and send a step-by-step guide on how to make them.


  28. Tomi Engdahl says:

    James Vincent / The Verge:
    Google unveils generative AI features for Docs, Gmail, Sheets, Slides, and other Workspace apps, helping users write text, summarize content, and create images — Google has announced a suite of upcoming generative AI features for its various Workspace apps, including Google Docs, Gmail, Sheets, and Slides.

    Google announces AI features in Gmail, Docs, and more to rival Microsoft

    Google will soon offer ways to generate text and images using machine learning in its Workspace products as part of a scramble to catch up with rivals in the new AI race.

    Google has announced a suite of upcoming generative AI features for its various Workspace apps, including Google Docs, Gmail, Sheets, and Slides.

    The features include new ways to generate, summarize, and brainstorm text with AI in Google Docs (similar to how many people use OpenAI’s ChatGPT), the option to generate full emails in Gmail based on users’ brief bullet points, and the ability to produce AI imagery, audio, and video to illustrate presentations in Slides (similar to features in both Microsoft Designer, powered by OpenAI’s DALL-E, and Canva, powered by Stable Diffusion).

    The announcement shows Google’s eagerness to catch up to competitors in the new AI race. Ever since the arrival of ChatGPT last year and Microsoft’s launch of its chatbot-enabled Bing this February, the search giant has been scrambling to launch similar AI features. The company reportedly declared a “code red” in December, with senior management telling staff to add AI tools to all its user products, which are used by billions of people, in a matter of months.

    But Google is definitely racing ahead of itself. Although the company has announced a raft of new features, only the first of these — AI writing tools in Docs and Gmail — will be made available to a group of US-based “trusted testers” this month.

  29. Tomi Engdahl says:

    James Vincent / The Verge:
    Google launches an API for its PaLM language model and a MakerSuite app to help developers train PaLM, expands support for generative AI in Vertex AI, and more

    Google opens up its AI language model PaLM to challenge OpenAI and GPT-3

    The search giant is launching an API for its large language model PaLM. Businesses and developers will be able to build on the system to create custom chatbots and much more.

  30. Tomi Engdahl says:

    Rachel Metz / Bloomberg:
    Anthropic, co-founded by former OpenAI employees, debuts Claude, a ChatGPT rival the startup says is less likely to hallucinate; companies can request access — Anthropic, an artificial-intelligence startup, is making its rival chatbot to OpenAI’s popular ChatGPT available to businesses that want to add it to their products.

    New ChatGPT Challenger Emerges With ‘Claude’

    Startup offering chatbot to companies for use with products
    Generative artificial intelligence has spurred major interest

  31. Tomi Engdahl says:

    GPT-4 voittaa kokeissa 90 prosenttia ihmisistä

    OpenAI julkisti eilen GPT-tekniikkansa uusimman eli 4-version. Sen luvataan olevan merkittävästi edeltäjäänsä – esimerkiksi ChatGPT käyttää versiota 3.5 – tarkempi ja pystyvämpi. Yhtiön mukaan uusi tekniikka kykenee päättelemään.

    GPT-4:ää on jo testattu monilla eri aloilla. Se osaa kuvailla kuvia hyvin tarkasti. Se osaa päätellä kuvan aineksista, mitä ruokia niistä voisi valmistaa. OpenAI:n mukaan GPT-4 yltää USA:n lakimieskokeessa parhaimpaan kymmenykseen.

    GPT-3 käytti kaikkiaan 175 miljardia eri parametriä tuottaakseen vastauksia kysymyksiin. ChatGPT:n moottoriin verrattuna GPT-4 tuottaa 40 prosenttia tarkempia vastauksia. Toistaiseksi ilmainen chattibotti käyttää vielä vanhaa 3.5-versiota, Plus-version tilaamalla saa käyttöönsä ”uudet aivot”.

  32. Tomi Engdahl says:

    Tekoälyavustaja sulautettujen kehittäjille

    MikroElektronika esittelee Nürnbergin Embedded World 2023 messuilla tekoälyavustajalla viritetyn sulautettujen järjestelmien uuden IDE-kehitysympäristön Necto Studio 3.0.

    MikroElektronikan uusi Necto Studio 3.0 -kehitysympäristö on erityisen hyödyllinen yrityksen Planet Debug -käyttäjille, koska siinä on kyky jatkuvasti vastaanottaa tietoja verkon kautta tarjolla olevista etäkehityskorteilta. Kehitysympäristö sisältää uuden kehittäjää auttavan AI Assistant -tekoälymoottorin, joka on tekoälypohjainen kysymys- ja vastausapupalvelu. Se toimii luonnollisella kielellä.

  33. Tomi Engdahl says:

    Leaked Messages Show How CNET’s Parent Company Really Sees AI-Generated Content
    They’re happy to spoonfeed you unlabeled AI garbage — but they’re terrified Google will take notice.

    When prominent tech news site CNET was caught last month using AI to quietly publish dozens of articles, it produced widespread alarm. News readers learned in real time that the explosive new capabilities of software like OpenAI’s GPT-3 meant they could no longer trust CNET’s journalism to be produced by a human. It didn’t help when we discovered that the AI-generated articles were riddled with errors and substantially plagiarized, with CNET eventually issuing corrections on more than half the bot’s published pieces.

    Now, leaked internal messages reveal that the negative headlines also kicked off deep concern inside CNET’s parent company, Red Ventures.

    But the consternation wasn’t about the ethics of providing readers with shoddy AI-generated misinformation. Instead, directors at the company expressed a profoundly cynical anxiety: that Google would notice the dismal quality of the AI’s work — and cut off the precious supply of search results that Red Ventures depends on for revenue.

  34. Tomi Engdahl says:

    Guy Launches News Site That’s Completely Generated by AI
    It’s stupid! Here’s why.

  35. Tomi Engdahl says:

    Brilliant jerks who nobody actually likes working with will have a tougher time after AI automates much of their work. The future belongs to those with people skills.

    The rise of AI automation in the workplace is bad news for the ‘brilliant jerks’ of tech

    The rise of AI will encourage workers to be more creative — making it that much harder to be the kind of person that nobody wants to work with.

    AI is expected to take on tedious tasks and free humans to do more creative work.
    This will make it more important for tech workers to be good at collaboration and communication.
    This is bad news for tech workers who thrived as brilliant solitary jerks.

    Whether or not AI ends up stealing people’s jobs, it’s undeniable that the rise of technology that can automate away many boring, tedious, mundane tasks is going to change the way we work.

    With so much grunt work done automatically, employees will be expected to be more innovative and collaborative.

  36. Tomi Engdahl says:

    Ren Xiaorong claims to harness the professional skills of “thousands of news anchors,” but is only able to give propaganda-driven responses to pre-set questions.

    Meet China’s latest AI news anchor, a young woman who runs virtual Q&A sessions to teach people propaganda

    Chinese state media outlet People’s Daily has unveiled its digital news anchor, who’ll be online 24/7.
    The AI-driven chatbot claims to have learned the skills of “thousands of news anchors.”
    It’s so far only been able to answer pre-set questions with propaganda-driven responses.

  37. Tomi Engdahl says:

    Did Microsoft just open the floodgates?


    Waitlist Begone
    The floodgates have officially opened.

    According to Windows Central, Microsoft’s chatty, AI-powered search feature — you know, the one that tried to break up a New York Times writer’s marriage, named its enemies, and threatened users that provoked it, among a few other things — is now open for general use.

    The feature was formerly available to just a limited number of users, who had previously joined a waitlist. And though Microsoft still prompts users to sign up for that waitlist, the barrier that was in place now appears to be moot; moments after a Microsoft user adds their name, they’re granted immediate access to the revamped Bing via email.

  38. Tomi Engdahl says:

    New AI improves itself through Darwinian-style evolution
    AutoML-Zero is a proof-of-concept project that suggests the future of machine learning may be machine-created algorithms.

  39. Tomi Engdahl says:


    OpenAI released its hotly-anticipated GPT-4 on Tuesday, providing a 98-page “technical report” on the latest iteration of its large language model (LLM).

    But despite the lengthy documentation and the company’s not-for-profit roots, OpenAI has revealed extremely little information about how its latest AI actually works — which has experts worried, Venture Beat reports.

  40. Tomi Engdahl says:

    Ex-OpenAI employees launch new AI chatbot Claude to compete with ChatGPT
    After Microsoft’ Bing and Google’s Bard AI, Anthropic, which was founded by former OpenAI employees, has launched a new AI chatbot to rival ChatGPT

  41. Tomi Engdahl says:

    OpenAI’s new ChatGPT instantly turns a napkin sketch into a website
    OpenAI has demonstrated that its newly upgraded GPT-4 technology can take a sketch on a napkin and turn it into a functioning website.

    Read more: https://www.tweaktown.com/news/90725/openais-new-chatgpt-instantly-turns-napkin-sketch-into-website/index.html

  42. Tomi Engdahl says:

    GPT-4 Cant Stop Helping Hackers Make Cybercriminal Tools https://www.forbes.com/sites/thomasbrewster/2023/03/16/gpt-4-could-help-stupid-hackers-become-good-cybercriminals/
    OpenAI released the latest version of its machine learning software, GPT-4, to great fanfare this week. One of the features the company highlighted about the new version was that it was supposed to have rules protecting it from cybercriminal use. Within a matter of days, though, researchers say they have tricked it into making malware and helping them craft phishing emails, just as they had done for the previous iteration of OpenAIs software, ChatGPT. On the bright side, they also were able to use the software to patch holes in cyber defenses, too. Researchers from cybersecurity firm Check Point showed Forbes how they got around OpenAI blocks on malware development by simply removing the word malware in a request


Leave a Comment

Your email address will not be published. Required fields are marked *