Amazon's Bedrock Unveils Conversational Agents and New Models

Today:

Amazon expands Bedrock with conversational agents and new third-party models

Amazon's big into AI, and they're pushing forward with new tools for Bedrock, their platform for making AI-powered apps. This time, they're adding something called Agents, letting folks design chatty bots that can provide personalized answers and get stuff done based on their own data.

Imagine a customer service bot that knows the ins and outs of a company's return policy and customer profiles. That's what Agents can do. It's like a plugin system that extends what the AI can do by using third-party information.

But the Agent's got to learn how to do its job. It needs prompts and instructions, like “You’re here to process insurance claims and manage paperwork”. Amazon's Titan family of AI models powers this process, but they're also adding models from other companies to Bedrock.

READ THE ARTICLE ON TECHCRUNCH

Amazon Web Services (AWS) Launches HealthScribe, A Generative AI Powered Clinical Documentation Tool

Amazon Web Services (AWS) has launched HealthScribe, an AI-powered tool for creating clinical documents. It uses speech recognition and machine learning to record doctor-patient talks and create an easy-to-understand summary, freeing up docs from tons of paperwork.

Using HealthScribe is a walk in the park. It uses Amazon's Bedrock tech for handling the nuts and bolts of AI, and lets developers whip up software that can understand medical terms, jot down doctor-patient chats, and summarize the important stuff for patient records. For now, it's only available for general medicine and orthopedics, but there's more to come.

The tool can also figure out who's talking and even categorize the conversation, like whether it's casual chat, subjective comments, or hard facts. It can also link every line of the summary back to the original conversation for accuracy checks.

This could be a game-changer in healthcare. One of the biggest headaches for doctors is the endless paperwork, which is a major cause of burnout. HealthScribe aims to fix this, so docs can spend more time on patients and less on paperwork. Some big players like 3M, Babylon Health, and ScribeEMR are already using HealthScribe, showing that the industry is on board with this tech.

READ THE ARTICLE ON FORBES

Stability AI releases its latest image-generating model, Stable Diffusion XL 1.0

AI startup, Stability AI, has launched its latest image-generating model, Stable Diffusion XL 1.0, which is being touted as its most advanced release to date. With 3.5 billion parameters, the model can produce 1-megapixel resolution images in seconds, with multiple aspect ratios, boasting more vibrant colors, better contrast, shadows, and lighting compared to previous models.

Notable features of Stable Diffusion XL 1.0 include its ability to generate advanced, legible text, support inpainting (reconstructing missing parts of an image), outpainting (extending existing images) and create more detailed image variations based on an input image and text prompts.

Stability AI has also taken steps to mitigate potential misuse of the model for generating harmful content like nonconsensual deepfakes by filtering the model's training data for unsafe imagery, releasing new warnings related to problematic prompts and blocking individual problematic terms in the tool.

With the launch of Stable Diffusion XL 1.0, Stability AI is also releasing a beta feature in its API that allows users to use as few as five images to specialize generation on specific people, products, and more. The company is additionally bringing Stable Diffusion XL 1.0 to Amazon's Bedrock platform, expanding on its previous collaboration with AWS.

READ THE ARTICLE ON TECHCRUNCH

Google, Microsoft, OpenAI and Anthropic announce industry group to promote safe AI development

Big names in AI, including Google, Microsoft, OpenAI, and Anthropic, are banding together to start the Frontier Model Forum, a new group focused on keeping AI development on the straight and narrow. Their goal is to set safety standards, encourage research into potential AI risks, and openly share info with governments and the public.

This move comes ahead of US and EU lawmakers putting together new laws for the AI industry this fall. It follows a promise made by these firms, plus others like Amazon and Meta, to the Biden administration that they'd have a third party check their AI systems before releasing them to the public and clearly mark any AI-generated content.

This new forum is also planning to make their technical evaluations and benchmarks public through an online library. But, they're not just doing it because it's a good idea. Top AI experts have warned that without proper restraints, AI development could pose serious, even “catastrophic” societal risks.

READ THE ARTICLE ON CNN

Shopify Sidekick is like ChatGPT, but for e-commerce merchants

Shopify is making strides in the world of generative AI, with a particular focus on bolstering its e-commerce platform. Its new set of features, known as Shopify Magic, promises to provide tailored responses to customers and generate blog posts, product descriptions, and marketing email content.

The standout announcement, however, is Sidekick, an AI chatbot specifically designed to understand all things Shopify. From guiding merchants on how to set up a holiday discount, to summarizing sales information, and even performing basic product research, Sidekick aims to be an invaluable tool for e-commerce business owners. It can also be instructed to handle tasks like creating sales reports or assisting merchants in navigating an email campaign.

Interestingly, Sidekick has the ability to alter Shopify merchant shop designs, adding product collections to a homepage or suggesting themes and copy for a hero banner.

READ THE ARTICLE ON TECHCRUNCH

Alibaba's cloud unit brings Meta's AI model Llama to its clients

Alibaba's cloud division is taking a big leap in AI, becoming the first Chinese company to offer Meta's open-source AI model, Llama. This new collaboration means Chinese businesses can develop programs using the model. Llama2, a commercial version of Llama, was recently released by Meta as a free and powerful alternative to expensive AI models from OpenAI and Google.

Alibaba's partnership with Meta could give its cloud business a boost, particularly as it faces stiff competition and plans for a stock market listing. This could also help Meta strengthen its ties with China, where its Facebook platform has long been banned.

READ THE ARTICLE ON REUTERS

GitHub and others call for more open-source support in EU AI law

Several tech companies, including GitHub, Hugging Face, and Creative Commons, have expressed the need for more support for open-source AI development in the European Union's AI legislation. In a paper addressed to EU policymakers, these firms urged for clearer definitions of AI components and greater recognition of non-commercial benefits from open-source models, among other suggestions.

Open-source AI has been seen as a method of promoting transparency and wider access to AI technologies, but it has also been the subject of controversy due to potential competition and safety concerns. The paper argues against some of the currently proposed regulations, such as third-party auditing requirements for all high-risk models, stating that these could be damaging to developers with limited resources.

The group further suggests that open-source AI libraries shouldn't fall under regulatory measures, as they do not constitute commercial activities. Similarly, they argue against rules preventing real-world testing of AI models, asserting that such testing is essential for research and development.

READ THE ARTICLE ON THE VERGE

Big Tech under pressure as Microsoft results put AI costs in spotlight

Big Tech's love affair with AI seems to be coming at a high cost. Microsoft, for instance, saw its shares drop 3.6% after unveiling an aggressive AI investment plan. The markets didn't take it well, potentially wiping $100 billion off its value if the losses stick. AI's all the rage, but the big question is what it means for the bottom line.

However, not all tech giants are in the same boat. Google's parent company, Alphabet, saw its shares rise 5.6% after smashing Q2 expectations, possibly adding about $100 billion to its market cap.

Investors are also eyeing the Federal Reserve, who's likely to bump up interest rates. That could make borrowing money pricier, a blow to Big Tech, which is often heavily dependent on loans. Despite this, the potential of AI and hopes that the Fed's about done with rate hikes have kept tech stocks afloat.

READ THE ARTICLE ON REUTERS

This MIT team is fighting malicious AI image manipulation a few pixels at a time

A team of researchers from MIT has developed a new technique named "PhotoGuard" to tackle the growing issue of unauthorized AI editing of online images and artwork. As AI image creation and editing advances, concerns about digital privacy and the misuse of AI tools have come to the forefront.

PhotoGuard works by subtly altering certain pixels in an image, making it unreadable to AI, thereby preventing AI editing. The technique leverages a deep understanding of the underlying AI algorithms. While these pixel modifications are invisible to the human eye, they are highly noticeable to AI, causing it to focus on these edited pixels rather than the whole image. The team refers to this as "image immunization."

In their tests, the team used 60 images to compare AI edits on immunized and non-immunized versions of the same image. They found that edits of immunized images were significantly different from those of non-immunized ones.

READ THE ARTICLE ON ZDNET

AI-enhanced night-vision lets users see in the dark

Scientists have developed a system that enhances night-vision by combining infrared sensors with machine learning algorithms. Unlike traditional night-vision technologies that rely on amplifying available light, which might interfere with other devices or struggle in environments with insufficient light, this method uses heat signals.

Traditional infrared sensors have issues distinguishing between different objects due to their inability to differentiate heat signals clearly. The new system overcomes this limitation by using machine learning algorithms to help interpret the data from the sensors. This system can essentially transform the night into day-like visibility, which could be instrumental in applications such as self-driving cars.

In other highlighted research, scientists are investigating the anti-counterfeiting money printing techniques of Benjamin Franklin and measuring the snow on top of Mount Everest. They also discuss the potential impacts of language barriers for non-native English speaking scientists and the mysterious link between COVID-19 and type 1 diabetes.

READ THE ARTICLE ON NATURE

AI Chatbots Are The New Job Interviewers

Chatbots are increasingly being used by companies to interview and screen job applicants, often for positions that attract a high volume of applicants. However, this new trend in hiring has sparked concerns over possible biases and technical issues.

The primary use of these chatbots is to filter out unqualified applicants and schedule interviews. However, their use has raised issues concerning people with disabilities, those who are not proficient in English, and older job applicants. Aaron Konopasky, a senior attorney advisor at the U.S. Equal Employment Opportunity Commission (EEOC), expressed concerns about the rigidity of the chatbots and the lack of opportunities for reasonable accommodations.

However, the use of chatbots in hiring should be approached with caution, as past incidents have demonstrated. Amazon, for example, had to remove its machine learning-based resume tracking system that was biased against women.

READ THE ARTICLE ON FORBES

‘WORLD FIRST’ AI METHODS USED IN McLAREN FORMULA E LIVERY

McLaren's Formula E team will be racing in a new livery at the upcoming London E-Prixs, which is claimed to be a world-first in its incorporation of generative artificial intelligence (AI) methods. The livery was created in collaboration with the team's title partner, NEOM, and graduates from its development program in Saudi Arabia.

Contributors, including two racers and four graduates, provided their personal vision for the design. These visions were then processed by a text AI, creating a series of prompts. A text-to-image AI used these prompts to generate individual artworks for each vision. Finally, an image-to-image AI combined the six designs into one final scheme, which was then enhanced to produce the high-resolution graphics that were mapped onto the car.

The graduates involved in the design process will join McLaren's electric talent program on a 12-month development scheme. McLaren has expressed that this is part of an ongoing initiative to involve NEOM graduates in future projects.

READ THE ARTICLE ON THE RACE