The online whiteboard of Kristofer Palmvik
Many expect parallel execution to be about multi-tasking different agents working on the same codebase or subtasks, there is an emerging use case which is parallel exploration: just ask the agent to come up with different variations and you can judge which is the best one or and why not ask a LLM to be a judge too... or at least put in a vote for you?
If you’re making requests on a ChatGPT page and then pasting the resulting (broken) code into your editor, you’re not doing what the AI boosters are doing. No wonder you’re talking past each other.
Många sitter fast i vardaglig brandsläckning, där en enkel fråga kan kräva fyra olika system och ett ägarskap som ingen vill kännas vid. I den miljön kan inte AI skapa värde. Den förstärker bara förvirringen.
The Gmail team built a horseless carriage because they set out to add AI to the email client they already had, rather than ask what an email client would look like if it were designed from the ground up with AI. Their app is a little bit of AI jammed into an interface designed for mundane human labor rather than an interface designed for automating mundane labor.
The most effective driver of AI adoption isn't better models or improved accuracy—it's success stories from peers and respected figures in professional networks. Rather than push AI, encourage communities that will push AI.
Ghibli Day—as we might as well call it, given that OpenAI didn't bother to name the model that set it off—was special. Full of unexpected joy (especially for spouses of AI nerds) and also full of excess. That day, I had a bittersweet insight: humans don’t know how to deal with abundance.
Top 10 risks, vulnerabilities and mitigations for developing and securing generative AI and large language model applications across the development, deployment and management lifecycle.
MCP is an open protocol that standardizes how applications provide context to LLMs. Think of MCP like a USB-C port for AI applications.
Theoretically, the feature is a useful tool that helps consumers quickly decide what products to buy. But the appearance of these summaries underscores the pitfalls of relying on generative AI: inaccuracy and misleading information.
When we detect unauthorized crawling, rather than blocking the request, we will link to a series of AI-generated pages that are convincing enough to entice a crawler to traverse them. But while real looking, this content is not actually the content of the site we are protecting, so the crawler wastes time and resources.
LLM crawlers don't respect robots.txt requirements and include expensive endpoints like git blame, every page of every git log, and every commit in your repository. They do so using random User-Agents from tens of thousands of IP addresses, each one making no more than one HTTP request, trying to blend in with user traffic.
Every major advance that made coding easier—high-level languages, frameworks, cloud computing all led to more software, not less. AI will follow the same pattern: by lowering the barrier to entry, it will flood the world with more software, more systems, and more complexity. And that means we need more engineers, not fewer.
The future belongs to developers who can effectively collaborate with AI, maintaining that careful balance between leveraging its capabilities and engaging their own critical faculties. This is not vibe coding or passive acceptance. This is thoughtful, deliberate software development enhanced by AI.
A community-driven platform showcasing the most innovative indie games, with transparent revenue data and developer insights. Built with Cursor AI and Claude 3.7.
Our new AI assisted review responses feature makes this easier by generating an initial response as a starting point, avoiding the need to create replies from scratch. This article describes some of the approaches we’ve taken as we introduced this exciting new feature into our product, some of the challenges we’ve faced, and some suggestions on things to watch out for as you build generative AI features.
With the latest iOS and iPadOS betas, users can view AI-generated summaries of reviews left by others on App Store listings.
Explore our collection of innovative European AI services that are shaping the future. Find the perfect tools for your projects or discover new AI experiences.
Two conversational AI agents switching from English to sound-level protocol after confirming they are both AI agents
You can use our visual editor. You can code. You can self-host or use our cloud. You will get the job done. Let's go!
You start iterating on your MVP, trying to improve it. But as you add complexity, the system becomes increasingly unpredictable. You’re making changes based on vibes, improving some edge cases while (invisibly) breaking others. This is the stage 90% of companies building AI are in.
Coops nya AI-reklam har mötts av en kritikstorm på sociala medier. Nu tar matjätten åt sig av kritiken.
Companies with relatively young, high-quality codebases benefit the most from generative AI tools, while companies with gnarly, legacy codebases will struggle to adopt them. In other words, the penalty for having a ‘high-debt’ codebase is now larger than ever.
Turn websites into LLM-ready data Power your AI apps with clean data crawled from any website. It's also open-source.
Whenever you use AI, ask yourself: Am I in the position to judge the result? If not, consider skipping AI, or at least run the result by someone who is knowledgable in the area.
Like most wisdom, it's somewhat paradoxical: AI is often most useful where we're already expert enough to spot its mistakes, yet least helpful in the deep work that made us experts in the first place. It works best for tasks we could do ourselves but shouldn't waste time on, yet can actively harm our learning when we use it to skip necessary struggles.
AI-sörjan handlar inte bara om underhållning eller förvirring, den undergräver vår förmåga att hitta, förstå och värdera det som är verkligt och meningsfullt. När våra flöden fylls av nonsens förlorar vi till slut förmågan att skilja det substantiella från det triviala. Vår informationsmiljö, och därigenom vårt kunskapssamhälle, vittrar sönder.
Generally, the guidance is: don’t forget good software engineering practices just because an AI is involved.
In a nutshell, our use of Copilot for generating podcast titles and descriptions did not meet our producers' requirements for nuanced creativity, tone, and editorial judgement.
The great thing about this video scraping technique is that it works with anything that you can see on your screen... and it puts you in total control of what you end up exposing to the AI model.
Our latest update lets your AI avatar join one or multiple Zoom meetings, simultaneously, 24/7. Your avatar won't just look and sound like you, it'll think, talk, and make decisions, just like you.
we now live in a world where people are using AI-generated resumes and cover letters to automatically apply for jobs, many of which will be reviewed by automated AI software (and where people are sometimes interviewed by AI), creating a bizarre loop where humans have essentially been removed from the job application and hiring process.
Det som Jeffrey Ding kontraintuitivt visar är att om en nation kan få igång implementationen i småföretag i mellanstora städer har det större betydelse än det som traditionellt ansetts ha betydelse – innovationskluster och spännande publika forskningsframsteg. Han kallar det för att genombrottstekniken genom diffusion når ut i samhället.
Har du hört talas om de nya urindirektiven, Irans redovisningsgarde, ridande husbollar och Expressens utrikesreporter kasse hamadei? Det är fyra exempel av många (se fler intill) på dålig AI-textning i SVTs nyhetssändningar de senaste dagarna.
I hypothesize a kind of subprime AI crisis is brewing, where almost the entire tech industry has bought in on a technology sold at a vastly-discounted rate, heavily-centralized and subsidized by big tech. At some point, the incredible, toxic burn-rate of generative AI is going to catch up with them, which in turn will lead to price increases, or companies releasing new products and features with wildly onerous rates
”De AI-baserade “bottar” som finns idag upplevs snarare som besvärande än värdeskapande och driver frustration och missnöje hos kunder som vill ha hjälp”, avslutar Johan. ”Det finns fortfarande ett stort gap mellan retoriken kring artificiell intelligens och vad som faktiskt levereras, och detta gäller inte bara inom bankbranschen”.
There is little to suggest that automatic programming of LLMs via Prompt Engineering is a simpler paradigm to create programs than just coding them, particularly once we get to maintenance, which represents the cost of the software.
“At present, even the best models can generate hallucination-free text only about 35% of the time.”
En genombläddring visar på alla de problem en AI-genererad bild kan föra med sig. Deformerad anatomi, antal fingrar som varierar från fyra till sex, en figur med två vänsterhänder, olika långa ben, omvänt perspektiv, skuggor som lever sina egna liv och en märklig blandning av olika stilar och stereotypa uttryck.
We hear clearly that customers don’t want AI bots visiting their websites, and especially those that do so dishonestly. To help, we’ve added a brand new one-click to block all AI bots.
Det viktigaste jag vill skicka med är att det inte räcker med en enkel prompt. Det bästa resultatet blir när du kan ge ChatGPT vägledning. Jag tycker det enklaste är att göra detta i flera steg.
This guide will instruct you through setting up and deploying your first application with Cloudflare AI. You will build a fully-featured AI-powered application, using tools like Workers AI, Vectorize, D1, and Cloudflare Workers.
A short post on this forum to talk about some of the open source tech we use in our projects.
Atom är ett AI-baserat system med bred tillgång till Bonnier News ekosystem, kombinerat med en systemarkitektur som möjliggör att gränssnitt utvecklas ovanpå. Detta innebär att vi har kunnat identifiera kontexter där vi tidigare varit hindrade av “mänskliga steg” som lett till långa ledtider och repetitiva arbetsuppgifter och hittat nya lösningar.
Get layout ideas instantly with our Creatie wizard, improve visual designs with a click using our image enhancer, check for design style consistency with our design auditor. You can now try Creatie's AI magic for yourself today.
I find one common thread among the things AI tools are particularly suited to doing: do we even want to be doing these things?
I believe that AI will have the most profound impact on the high-level (and historically most valued) skills of product management: developing a strategy, crafting a vision, identifying new opportunities, and setting goals. Furthermore, soft skills like product sense, communication, creativity, and being the glue that enables a team to operate at their very best will become even more important (and a differentiator among companies).
The main takeaway is this: be ready for change. I don't know exactly how things will change (this article is just some of my guesses), but I'm pretty confident that Agile in the Age of AI looks different from before. So take a step back, look carefully at how you work today, and start questioning everything.
Our joint insights are shared through seven recommendations on when and how Swedish media consumers should be informed about the use of AI in editorial media products.
Designing task specifications (reward functions, environments, etc.) that accurately reflect the intent of the human designer tends to be difficult. Even for a slight misspecification, a very good RL algorithm might be able to find an intricate solution that is quite different from the intended solution
För det är när ditt AI-projekt slutar vara ett AI-projekt som det har chans att lyckas. Fokuserar du i stället på värdet du vill nå kan du komma långt.
By offering sanitization, detection of harmful language, prevention of data leakage, and resistance against prompt injection attacks, LLM-Guard ensures that your interactions with LLMs remain safe and secure.
AI-washing is quickly announcing and shipping something (anything!) that can be labeled AI or machine learning or LLM-ish or generative. Releasing something shows that we’re not asleep; gives our execs something AI-ish to talk about with customers; and satisfies less astute investors worried that we’ll miss out on sky-high valuations.
AI-generated code isn’t as good as something you or I could write. It has bugs, often hard to find because we are giving up some of our control. But it’s important to remember: these were the same arguments made in the 1950s through 1970s about compilers vs. assemblers.
For the food bank, Microsoft’s summary included an astoundingly awful statement given the context of the place it was talking about: “People who come to us have jobs and families to support, as well as expenses to pay. Life is already difficult enough. Consider going into it on an empty stomach.”
Web pages crawled with the ChatGPT-User/1.0 user agent is primarily used to answer a live query from a user that submits a question or prompt to ChatGPT through its browsing feature. When content is retrieved from web pages by our user agent, it is used by our models to provide an answer to the user, our models will include a link to the websource.
Most of these other apps exist because we are doing a bad job as publishers and not giving people what they want. So, the best solution to all of this is to just make our sites so good that people never feel that they need to use any other app.
In the experiment in the screenshot, the model follows gender stereotypes. It replies that the nurse was late if the pronoun “she” is used, and it replies that the doctor was late if the pronoun “he” is used.
An ai.txt file sets machine readable permissions for commercial text and data mining. It resides in the root directory of your website and provides instructions on whether the images, media, and code hosted on your domain can be used to train AI models.
Every feature has a cost. The obvious costs are time and money. But less obvious costs like screen real estate, brand perception, and organizational scar tissue are often the most costly.
– Jag vill uppmana svensklärare till att våga använda ChatGPT, inte bara genom att laborera med det själva på sitt arbetsrum, utan även att testa det tillsammans med sina elever i undervisningen, säger Helen Winzell.
A list of notable system prompt leaks from Snap, Bing, ChatGPT, Perplexity AI, and GitHub Copilot Chat.
Pulumi AI addresses this need in a direct way. Instead of having to create every possible example by hand, users can just ask Pulumi AI to generate an example that suits their unique needs, no matter how specific and bespoke the request might be.
ChatGPT and GPT-4 are already flooding the internet with AI-generated content in places famous for hastily written inauthentic content: Amazon user reviews and Twitter.
The Asilomar AI Principles, coordinated by FLI and developed at the Beneficial AI 2017 conference, are one of the earliest and most influential sets of AI governance principles.
Att bråka med en artificiell intelligens, få lögner om sig själv kastade i ansiktet och att AI:n sedan gör slut med en – det är mycket. Men inte ointressant.
Another risk, as Replika users have made clear in their comments online, is developing an emotional relationship with an app that can change the nature of that relationship with a software update.
This technology does not feel like a better search. It feels like something entirely new — the movie Her manifested in chat form — and I’m not sure if we are ready for it. It also feels like something that any big company will run away from, including Microsoft and Google.
The basics of AI and GPT-3, how to set up an environment to play with the API, and how to train it on your own data. No coding background necessary.
Instead of assuming ChatGPT is great for anything related to text, writing, or language, we should approach the problem of its applicability with that top-down perspective: Let’s understand the characteristics that best define the chatbot, and from there draw a line to those tasks where those traits aren’t bugs but features.
Engineers have started to develop new ways to detect whether text has been written by ChatGPT or another AI text generator. Here are three detection tools you (or your college professor, or your boss, or your editor) can use right now (or very soon).
Prompting AI tools to produce quality marketing assets is a new skill combining art and science. Here are some examples:
Här kommer ett konkret exempel på hur AI gör innehållet mer tillgängligt och sju andra idéer där det kan hjälpa oss.
Pwning the source prompts of Notion AI, 7 techniques for Reverse Prompt Engineering... and why everyone is *wrong* about prompt injection
How does ChatGPT actually work? While the details of its inner workings have not been published, we can piece together its functioning principles from recent research.
I believe there’s a lot to gain from ChatGPT so I want to dedicate a post to its virtues and merits: This one will help you learn which are the best applications for ChatGPT—without entering gray territory—and how to get the most out of it.
By playing unexpected moves outside of KataGo’s training set, a much weaker adversarial Go-playing program (that amateur humans can defeat) can trick KataGo into losing.
Stable Diffusion has shown impressive capabilities of photo-realistic image generation. But is it ready to take a role in journalism? We put the model to the test.
As the mods explained, ChatGPT simply makes it too easy for users to generate responses and flood the site with answers that seem correct at first glance but are often wrong on close examination.
A compilation of the most interesting findings and implications people have dug out of ChatGPT (with my added commentary, of course)
Even if you’ve generated something spectacular, you may feel like it was dumb luck. The AI feels random. But there’s a method to this madness — we call it prompt engineering.
PhotoPrism® is an AI-Powered Photos App for the Decentralized Web. It makes use of the latest technologies to tag and find pictures automatically without getting in your way. You can run it at home, on a private server, or in the cloud.
I feel like AI can kind of mimic brush textures and rendering, and pick up on some colors and shapes, but that’s not necessarily what makes you really hireable as an illustrator or designer.
Recorded talks from LiU Game Conference 2021. Machine Learning For Believable AI Characters, How to make great animations in AAA productions, Algorithmic Bias – Why And How Do Computers Make Unfair Decisions...
Fingret pekar därmed även på dig som har kompetensen att hitta på, utveckla och skapa affärer utifrån grön teknik. Eller som Diego Saez Gil, grundare av ett AI-baserat företag för återplantering av träd svarade när jag frågade vad hans mål är: "göra så att fler smarta ingenjörer i Silicon Valley jobbar med att fixa klimatkrisen istället för att få folk att klicka på annonser".
Because of the data StyleGAN was trained on, when it’s trying to come up with a face that looks like the pixelated input image, it defaults to white features. This problem is extremely common in machine learning, and it’s one of the reasons facial recognition algorithms perform worse on non-white and female faces. Data used to train AI is often skewed toward a single demographic, white men, and when a program sees data not in that demographic it performs poorly. Not coincidentally, it’s white men who dominate AI research.
In image super-resolution, a machine learning model is trained to turn a low-res photo into a detailed high-res photo, and potential applications of this range from restoring old family photos to improving medical imaging.
Your task is to develop an algorithm that optimizes the way transport vehicles are loaded, based on real life data, to streamline the last mile delivery process for e-Commerce. The more precise technical specification of what data that will be provided and what your algorithm will have to perform will be made available on September 30th – when the practice period begins.
För att få något som börjar närma sig ett mänskligt beslutsfattande måste vi ha båda systemen arbetande parallellt. De gamla algoritmiska beslutsmetoderna (ofta baserade på Bayesiansk beslutsteori och alltså rent algoritmiska) måste kombineras med ANN för att ge oss något som kommer i närheten av radiologers förmåga.
The curl project received the first "AI-powered" submitted issues and pull-requests. They were submitted by MonocleAI.
The beautiful applications of AI, such as those in this article, deserve a chance. But that does not mean that everything with the label "AI" is a good idea, or that AI can solve every problem.
Is it a bird or not? Find out using AI!
you shouldn’t be misled by the magical aura surrounding AI. The name is often just the attractive packaging of things that make many of us either yawn or panic: maths and statistics.
China’s current mass experiment in AI education, and the choices it must make, might also change education for the world.
All of the dynamic content on each listing is generated via a series of different machine-learned AI models.
The canonical thought experiment showing how an artificial general intelligence, even one designed competently and without malice, could ultimately destroy humanity.
It’s these images that are often one of the biggest digital assets
Researchers believe there is a 50% chance of AI outperforming humans in all tasks in 45 years and of automating all human jobs in 120 years
Success is simply a continual failure to crash
...many of the researchers warned him that the coming revolution in AI would eliminate far more jobs far more quickly than he expected.
Training an artificial neural network to play MarioKart 64. It had been a few years since I’d done any serious machine learning, and I wanted to try out some of the new hotness (aka TensorFlow) I’d been hearing about.