- NATURAL 20
- Posts
- AI Apps Making $20,000+ per month with 1 person teams.
AI Apps Making $20,000+ per month with 1 person teams.
Learn about groundbreaking AI applications in business, and find out how to turn AI insights into profitable business strategies
Wes Roth
December 04, 2023
Today:
AI Apps Making $20,000+ per month with 1 person teams.
2023 was all about the buzz, showing off what AI might do. But now, it's game time – real uses, real money. The video talked about how AI can run your whole business while you snooze, handling customer service, making cash, the whole nine yards. They shared examples of businesses raking in big bucks, like up to $8 million a year, with just one person at the helm, thanks to AI.
AI's only gonna get bigger. Goldman Sachs says 2024 is the year AI really hits its stride. So, if you've got an idea, now's the time to paddle out and catch that wave.
OpenAI Agreed to Buy $51 Million of AI Chips From a Startup Backed by CEO Sam Altman
OpenAI was planning to spend a whopping $51 million on some fancy AI chips from a startup called Rain AI. The twist? Sam Altman had some skin in the game - he'd put over a million bucks into Rain AI himself. This company's just a stone's throw away from OpenAI in San Francisco and they're cooking up a chip that's supposed to think like a human brain.
Rain AI's got big plans for their chips, which they say could outdo the usual AI workhorses from Nvidia. But they're hitting some bumps, like leadership changes and some heat from the U.S. government over a Saudi Arabia-linked investment in the company. This cash influx from Saudi was a no-go for the U.S. folks worried about national security, so it got axed.
Rain AI's aiming to make chips not just for big data centers but for gadgets on the go like phones and drones. They're promising more power and efficiency, which could be a game-changer for AI. But it's still up in the air how OpenAI would use these chips.
AstraZeneca ties up with AI biologics company to develop cancer drug
AstraZeneca just teamed up with a brainy AI outfit called Absci Corporation. They're throwing down up to $247 million to cook up a cancer-fighting antibody. It's all about using Absci's smart AI to dig into proteins and find a killer therapy for cancer.
AstraZeneca is all about making new drugs without the crazy high costs. They're one of many big pharma companies linking up with these young AI whizzes to get the job done faster and cheaper.
These Absci folks are based in Washington, with a lab in New York. They're doing something pretty neat – measuring zillions of protein interactions to train their AI. Then, they use this to whip up antibodies that can spot and attack foreign stuff in our bodies.
The exact cancer they're targeting is still under wraps. But AstraZeneca's got big dreams of ditching old-school chemo for new, targeted drugs. They've already had some wins with lung and breast cancer treatments.
OpenAI Gives Employees Extra Month to Opt Into Plan to Sell Shares
OpenAI's bigwigs have decided to stick with their plan to let employees sell their shares through what’s known as a tender offer. They're giving their team an extra month, until January 5th, to make up their minds about selling.
Back in October, word got out that OpenAI might sell shares in a deal pegging the company at a whopping $86 billion. But things got shaky earlier this month. There was a bit of drama with the CEO, Sam Altman – he got fired and then, in a surprising twist, got his job back pronto.
This back-and-forth caused some jitters among investors, with a few deciding to bail. But after Altman was back, enough investors stepped up to fill the gap left by those who walked away. OpenAI's spokesperson said that despite the drama, no investors actually dropped out of the tender offer.
Why's this a big deal? An $86 billion valuation rockets OpenAI to the top of the startup world. It's been only a year since they dropped ChatGPT and shook up Silicon Valley, making them the hotshot in the AI scene.
MediaTek's AI-Powered Chips Aim To Make Your Next Phone Very Personal
MediaTek, a big player in the phone chip game, is all about bringing AI directly into your phone. They're talking faster, more personal responses, without having to rely on cloud computing. Think about having a phone that can jot down and summarize your meetings on the spot – handy, right? Vivo's already working on this for a phone release next year.
MediaTek's got two new chips, the Dimensity 9300 for fancy phones and the Dimensity 8300 for the budget-friendly ones. Both are AI-smart, but the 9300 is the real show-off with some serious image-generating speed.
Now, there's no clear-cut way to measure how good these AI chips are yet, but MediaTek's working on it. They're also thinking about how to keep things real in an AI-heavy world, like adding digital watermarks to AI-tweaked photos.
MediaTek's thinking cars too, like using AI to spot stuff on the road or remind you to close your car windows. They're also upping their game with Wi-Fi 7 and working on the next big thing, Wi-Fi 8, to connect everything better. Plus, they're diving into the AR/VR world, teaming up with big names like Meta for custom chips. It's all about making tech that fits each brand's needs.
OpenHermes 2.5 Mistral 7B beats Deepseek 67B and Qwen 72B on AGIEVal, and other 13B and 7B models!
OpenHermes 2.5 Mistral 7B, a hot topic in the LLM world, is outperforming big guns like Deepseek 67B and Qwen 72B in the AI evaluation game. It's not just a one-hit wonder either; it's leading the pack in both 7B and 13B categories.
It's a beefed-up version of Mistral 7B, souped-up with over a million entries of top-notch data, including stuff from GPT-4. It's acing tests like TruthfulQA and GPT4All.
To see how cool it is, they put it through three tests: explaining democracy to a kid, solving a basic train speed problem, and giving the lowdown on the French Revolution. And guess what? It nailed them all with answers that are on point and easy to get. They also checked out its sibling, OpenHermes-2.5-neural-chat-7B-V3–1–7B. This one's a combo of OpenHermes and Intel's tech. It's also acing the leaderboard in its categories.
The setup for OpenHermes 2.5 Mistral 7B is a bit of a chore, involving some tech wizardry with Python and other tools. But once it's up, you can ask it anything.
Hewlett Packard Enterprise and Nvidia Extend Their AI Collaboration
Hewlett Packard Enterprise (HPE) and Nvidia are teaming up big time to ramp up HPE's AI game. They're saying this partnership is gonna turn businesses into AI whizzes. Nvidia's pumped, believing this move will turbocharge what generative AI can do.
HPE's diving deeper into generative AI with Nvidia's help. They're cooking up a new line of products that mix HPE's cloud, supercomputing, and AI smarts. Their CEO, Antonio Neri, is talking about how AI needs a whole new tech approach, especially for handling data and crunching numbers. They're aiming to let businesses use their own data to build AI models without risking security.
iA Writer can now track what you or ChatGPT wrote
The latest version of iA Writer, a minimalist writing app, has rolled out a cool new tool. It lets you see the difference between what you wrote and what an AI like ChatGPT spits out. Your own words stay in black, while anything the AI writes shows up grey. This helps you keep track of your original thoughts versus AI suggestions.
iA Writer isn't just throwing this feature out there for kicks. They've got a whole philosophy about it. They don't want AI to ghostwrite for you and steal your thunder. Instead, they think of AI as a buddy you chat with to spark new ideas and improve your writing game.
This feature needs a bit of elbow grease from you. During a test, it was noticed that the app only recognizes AI text if you copy-paste it with the original prompt you gave the AI. If you just copy the AI's reply, you gotta manually tag it as AI text.
For now, this fancy tool is only on the Mac, iOS, and iPadOS versions of iA Writer. They're planning to bring it to Windows and Android soon. Plus, they've put the specs on Github and are open to sharing this feature with other apps, hoping it becomes a new standard in writing software.
Error-Detection Tool Makes AI Mistakes Easy to Spot
Usually, neural networks look at data and make guesses, like picking out faces in photos. But how they learn to make these guesses is kind of a mystery. The new study from Purdue takes a different approach. Instead of trying to follow the AI's thought process for each guess, they look at the big picture, mapping out how the AI sees all the data in a database.
They tested this out on a huge pile of images, like 1.3 million, and figured out a way to spot pictures that the AI might confuse for being in more than one category. They used some fancy math from a field called topology to map out how the AI links each picture to different categories.
These maps they make show groups of images as dots, color-coded by what the AI thinks they are. Most of the time, you see clusters of dots in one color. But when there's a mix-up, you see dots of different colors overlapping. That's where the AI might be getting confused.
By looking at these maps, you can tell where the AI has trouble telling two categories apart. This method even helped them figure out why an AI kept mixing up car images with cassette players—it turned out the car photos had tags for stereo equipment!
EY claims success in using AI to find audit frauds
EY, one of the big accounting firms, tried out this AI system to catch fraud in some UK companies. Guess what? It worked pretty well! They found shady stuff in two out of the first ten companies they checked. This is big news because it shows that AI might be a game-changer in finding bad business practices and making auditors' lives easier.
The cool thing about AI is that it learns on the go, getting better at spotting fishy business the more it sees. This could really help auditors, especially since there's a crunch to find and train enough people for the job.
EY's AI tool was trained on a whole bunch of fraud cases, both public and from their own files. It looks for the sneaky moves used to hide fraud, not just the obvious red flags. But other firms, like KPMG, are a bit skeptical about AI catching the really slick fraudsters.
This faux AI chatbot will judge your music taste and make you laugh (hopefully)
Spotify just rolled out their 2023 Wrapped, showing everyone's top jams for the year. Now, there's this sassy fake AI on The Pudding's website, "How Bad Is Your Streaming Music?". It's all about judging your tunes on Spotify or Apple Music. The thing dishes out some serious burns and asks questions to figure out how your music stacks up.
So, you wanna try it? Head to their site, hit "Find Out", and pick whether you're a Spotify or Apple Music fan. Don't sweat it; they won't mess with your account. Log in, and get ready for some cheeky questions about your music choices. Heads up, though: the AI's got a bit of a potty mouth.
After you've had your laugh, the site spits out a score out of 100. Don't expect to be blown away – the scores are pretty stingy. Some folks hit a snag with a blank screen at the end, but it's probably just the site being swamped. Even if you don't see your final score, it's a good time.
AI system self-organises to develop features of brains of complex organisms
Researchers at the University of Cambridge explored how artificial neural systems can self-organize and develop features similar to those found in human brains. They used a simplified artificial system that mimics the brain's functions, with computational nodes replacing real neurons. The key insight is that when physical constraints were applied, making it harder for distant nodes to communicate, the artificial system began to exhibit characteristics seen in human brains.
When given a maze navigation task, the artificial system learned by adjusting the strength of connections between nodes, similar to how human brain cells change connections during learning. Notably, the system developed hub nodes, much like the human brain's highly connected regions.
What's intriguing is that individual nodes in the artificial system developed flexible coding schemes, allowing them to encode multiple properties of the task, a feature seen in complex biological brains. This study provides insights into why brains, including those of humans, are organized the way they are.
Martian’s tool automatically switches between LLMs to reduce costs
Shriyash Upadhyay and Etan Ginsberg, AI researchers from the University of Pennsylvania, believe that many big AI companies are more focused on outdoing each other than on doing basic research. They think this is because when these companies get a lot of money, most of it goes into competing rather than studying the basics.
To address this, Upadhyay and Ginsberg founded a company called Martian. Martian's goal is to make AI research profitable by emphasizing interpretability over sheer power. Martian recently came out of hiding with $9 million in funding from investors like NEA, Prosus Ventures, Carya Venture Partners, and General Catalyst. They plan to use the money for product development, researching how AI models work, and expanding their team.
Martian's first product is a "model router" that figures out which large language model (LLM) is best for a given task. For example, if you need help with a math problem, it'll send your request to the LLM best at math. This can save money compared to relying on a single high-end LLM like GPT-4, which can be expensive. Companies like Permutable.ai spend over $1 million a year on such models.
Can artificial intelligence improve life science? As much as life science can improve AI, researchers say
Michael Q. Zhang, a professor at the University of Texas, Dallas, highlights how AI can revolutionize science and technology, benefiting daily life and social interactions. He encourages discussions on this topic.
Xuegong Zhang and his team propose a concept called Digital Life Systems (dLife), aiming to integrate AI into biology and medicine comprehensively. They envision dLife creating digital replicas of systems, including individual human bodies, for quicker and more accurate insights into treatments.
Gangqing Hu and his team introduce the OPTIMAL model, which uses ChatGPT to aid beginners in bioinformatics. This model enhances coding skills and critical thinking for students, extending beyond classrooms.
Dong Xu sees potential in ChatGPT for bioinformatics and tutoring, while Jianfeng Feng explores the quest for artificial general intelligence (AGI), which goes beyond current AI capabilities.Feng believes our brain's probabilistic nature sets it apart from computers. His research aims to simulate the entire human brain, a massive interdisciplinary endeavor.
Everything you need to know to create a custom GPT: Unpacking OpenAI’s new GPT feature
This article is a beginner's guide to creating customized GPTs, focusing on OpenAI's recent DevDay announcement. It explains what GPTs are, how to use and enable them, and the benefits they offer.
A GPT, or customized version of ChatGPT, lets users personalize their AI experience without any coding. You can add features like DALL-E (image generation), Browsing with Bing, Code Interpreter, custom instructions, file upload, and plugins. OpenAI plans to introduce a GPT store for creators to earn revenue from their GPTs.
Benefits of using GPTs include creating task-specific GPTs, a thriving community and store for sharing and monetizing GPTs, and a focus on privacy and safety with a review process and builder profiles.
What'd you think of today's edition? |
What are MOST interested in learning about AI?What stories or resources will be most interesting for you to hear about? |
Reply