Introduction
It's been an incredible week for AI with some game-changing news. In today's blog, we'll delve into some of the most intriguing AI stories, covering Microsoft, Google, OpenAI, Alibaba, Samsung, and more. We'll discuss Microsoft's new project called Algorithm of Thoughts (AOT), Google's feature called Synth ID, the ongoing debate between OpenAI and Google, OpenAI's Chat GPT Enterprise, Alibaba's innovative AI model Quenvl, Samsung's AI-driven food service Whisk, an AI camera in Cornwall aiming to improve road safety, Walmart's internal AI tool My Assistant, and finally, a remarkable AI system from Google and Osmo that can smell just like humans.
Microsoft's Algorithm of Thoughts
Microsoft is working on a new project called Algorithm of Thoughts (AOT). The idea is to make AI-like chat GPT smarter and better at solving problems. AOT uses a set of rules to guide the AI in making decisions. It's like playing a game where the AI has to follow steps to win. For example, if the AI needs to summarize a long article, AOT will direct it to find the main points first, put them in simpler terms, and then make sure everything makes sense. Microsoft says this method is so good they used it to design a new computer chip specialized for AI tasks. They even made an upgraded version of chat GPT using AOT that can handle more complex chats with people. The benefits of AOT could be significant. It could speed up how computer chips are made and make AI-like chatbots and digital assistants interact more naturally with people. It could also help with creative tasks like writing. AOT is seen as a big step forward in AI.
Google's Synth ID
Google has introduced a new feature called Synth ID to fight misinformation. This feature places an invisible permanent watermark on images created by Imogen, Google's own text to image generator. The watermark stays even if the image is edited, helping identify whether an image is AI-generated. There are three levels of certainty: detected, not detected, and possibly detected. The feature is currently in beta and available for some users of Vertex AI, Google's platform for generative AI. Google hopes to improve Synth ID and may include it in other products. This move comes as fake and edited images are becoming a big problem online. Google is joining other companies in the effort to make it easier to tell what's real and what's not.
The Debate: OpenAI vs. Google
The research firm Semi-Analysis recently published an article claiming that Google's upcoming Gemini AI model will be much more powerful than OpenAI's GPT4. The authors of the report, Dylan Patel and Daniel Nishbal, say that Google has access to much better computer chips, making Gemini perform better in terms of computer calculations or flops. In response, OpenAI's CEO, Sam Altman, questioned the source of the report and hinted that the information was likely flattering because it came from Google. Online discussions are now focused on whether having better chips actually makes an AI model better. Some argue that quality isn't just about computational power, but also includes things like training processes and data quality. Other companies like Meta have also entered the field of large language models, and with Gemini releasing later this year, the competition is heating up.
OpenAI's Chat GPT Enterprise
OpenAI has launched a new version of its chatbot called Chat GPT Enterprise. It's designed for businesses and is packed with features to keep data safe. Unlike the regular version, it doesn't use data for training and is SOC2 compliant, which is a high standard for data security. Managers can also control who has access to it. Performance-wise, this new version is built on the latest GPT4 technology and can handle a lot of usage. It comes with a special code interpreter feature that helps understand data by creating charts or doing complicated analyses. It can even be customized using chat templates or API credits to fit specific tasks like customer service. Big companies like Canva and PWC are already using it, and OpenAI plans to add more features like code writing and image recognition. This new version of Chat GPT is seen as a game-changer for how businesses use AI.
Alibaba's AI Model Quenvl
Alibaba has created a new AI model called Quenvl that can understand both pictures and words in English and Chinese. This AI model is really good at tasks like recognizing what's in a photo and answering questions about it. It can even tell you where a picture was taken and can handle multiple questions in a row. Quenvl is useful for many things, like answering questions based on images and documents, creating captions for photos, and recognizing small details in pictures. Right now, anyone can use Quenvl for free or for business on Alibaba's model scope platform. Alibaba says this is the first AI model like this in China and one of only a few in the world. The introduction of Quenvl by Alibaba is seen as a significant advancement in AI technology.
Samsung's AI-Driven Food Service Whisk
Samsung has introduced a new service called Whisk that uses AI to help you with everything related to food. Available in over 80 countries, it picks recipes that you'll like, adjusts them for serving size, and even helps you shop for ingredients. Whisk uses AI to personalize your food choices and recipes based on your preferences, dietary needs, and available ingredients. For example, Whisk can suggest recipes that match your taste, budget, and health goals. It can also adjust the recipes according to the number of servings, cooking time, and nutritional value. It can work with your online grocery store to make your shopping list and even connect to your smart fridge and oven to keep track of what you have and how your cooking is going. Samsung says Whisk can save you time, reduce waste, and help you make tasty and healthy meals. Whisk is seen as a helpful app for making cooking easier and more enjoyable.
AI Camera in Cornwall for Road Safety
Cornwall, a county in England, has recently installed an AI camera that can detect traffic violations such as mobile phone use and seat belt offenses. The camera, called Lane Watch AI, is developed by a company called ZenGrab. Lane Watch AI is a smart camera that can capture high-quality images of vehicles and drivers in real-time. It uses AI to analyze the images and identify any violations or anomalies. For example, it can recognize if a driver is holding a phone, wearing headphones, or not wearing a seat belt. It can also detect if a vehicle is speeding, tailgating, or changing lanes without signaling. It then sends the images and the details of the violations to the authorities, who can issue fines or warnings to the offenders. ZenGrab says that Lane Watch AI can help improve road safety and reduce accidents by deterring dangerous driving behaviors. The camera is currently being tested in Cornwall as part of a pilot project and is located on a busy road near Newquay airport, monitoring around 20,000 vehicles per day. The project has been running for six months and has already recorded more than 5,000 violations. Lane Watch AI has received positive feedback from the local community and the police. ZenGrab hopes that their camera will be adopted by more regions and countries in the future.
Walmart's AI Tool My Assistant
Walmart is giving its corporate workers an AI tool called My Assistant to help them get work done faster. The tool can quickly summarize documents and help with projects. The aim is to free workers from boring repetitive tasks so they can focus on making the customer experience better. While other big companies like Amazon and Apple haven't fully adopted this tech in their offices, Walmart is going all in. Walmart is already using AI in other ways too. They have chatbots for customer service and a voice assistant in stores to help employees find items. They also have AI cameras at self-checkout to prevent theft. The CEO said they plan to use more AI to understand customers better and improve how they manage stock. Walmart believes the benefits of using AI like this are greater than any downsides and that both people and technology are needed to achieve their goals.
Google and Osmo's AI System That Can Smell
Scientists from Google Research and a startup called Osmo have made an AI system that can smell just like humans do. They used a neural network to teach the AI how to describe the smell of about 5,000 different chemicals. The AI did this by looking at the chemical structures and linking them to smells like fishy or woody. This AI is really accurate and often even better than people at describing smells. The team made a principal odor map (POM) which the AI uses to guess the smell of new chemicals. The guesses are usually really close to what people say. This AI could help us understand more about how our brain works with smells. It could also be used to create new scents for products. The development of this AI system by Google and Osmo is seen as a significant breakthrough in sensory AI.
Conclusion
In this blog, we covered some of the latest AI news, including Microsoft's Algorithm of Thoughts, Google's Synth ID, the debate between OpenAI and Google, OpenAI's Chat GPT Enterprise, Alibaba's AI model Quenvl, Samsung's AI-driven food service Whisk, an AI camera in Cornwall for road safety, Walmart's AI tool My Assistant, and Google and Osmo's AI system that can smell like humans. These advancements in AI technology have the potential to revolutionize various industries and improve our daily lives. From smarter chatbots to enhanced road safety, AI continues to push boundaries and open up new possibilities. The future of AI looks promising, and we can expect more exciting developments in the coming years.
.jpg)
0 Comments