According to a recent story from The Information, OpenAI’s next-generation Orion model of ChatGPT, which is both rumored and denied to be arriving before the end of the year, might not be everything that has been hailed when it does.

According to the report, which quotes unnamed OpenAI staff, the Orion model has shown a “far smaller” improvement over its predecessor, GPT-4, than GPT-4 did over GPT-3. According to the same individuals, Orion “isn’t reliably better than its predecessor [GPT-4] in handling certain tasks,” particularly coding applications, even though the new model is noticeably better at general language functions like creating emails or summarizing documents.

According to the Information report, one of the main reasons for the new model’s negligible improvements is the “dwindling supply of high-quality text and other data” that may train new models. Simply put, the AI sector is rapidly approaching a training data bottleneck after removing the simple sources of social media data from websites like X, Facebook, and YouTube (the latter twice). Pre-release training is being slowed significantly because of these organizations’ growing inability to identify the kinds of complex coding problems that will help them push their models beyond their existing limits.

This decreased training effectiveness has significant business and environmental ramifications. The amount of energy, water, and other resources is predicted to increase sixfold over the next ten years as frontier-class LLMs expand and continue to push their parameter counts into the high trillions. Because of this, the country’s current power infrastructure cannot supply the power required for their expanding networks of artificial intelligence data centers. For this reason, Google is purchasing the output of seven nuclear reactors, AWS is purchasing a 960 MW plant, and Microsoft is attempting to restart Three Mile Island.

To get around the shortage of suitable training data, OpenAI has established a “foundations team.” These methods might entail the use of artificial training data, like that produced by Nvidia’s Nemotron family of models. The group is also investigating ways to enhance the model’s performance after training.
​
Initially believed to be the code name for OpenAI’s GPT-5, Orion is now expected to arrive in 2025. It’s unclear if we’ll have enough electricity to see it through to completion without causing our municipal electrical grids to brown out.

A novel AI model known as CHIEF (Clinical Histopathology Imaging Evaluation Foundation), developed by researchers at Harvard Medical School, has been very accurate in diagnosing and forecasting the course of certain cancer types.

The study claims that CHIEF works better than current AI systems, reaching up to “96% accuracy” in the diagnosis of 19 distinct cancer kinds. The researchers compare CHIEF’s adaptability to that of ChatGPT, a language model that has drawn a lot of interest because of its capacity to handle a variety of jobs.

Instead of taking the generalist approach found in more conventional models like GPT-4V or LlaVA, CHIEF is essentially a highly specialized AI vision model—one that can comprehend visual inputs—trained to be extremely detailed in images of cancer cells.

Therefore, CHIEF was trained on a large multimodal dataset that included 15 million unlabeled photos and 60,000 whole-slide images of tissues from 19 different anatomical sites, rather than being trained to recognize generic elements like “cats” or “oranges.”

According to the study, “CHIEF extracted microscopic representations useful for cancer cell detection, tumor origin identification, molecular profile characterization, and prognostic prediction through pretraining on 44 terabytes of high-resolution pathology imaging datasets.”

The strategy appears to be more effective than expected. According to research senior author Kun-Hsing Yu, “our goal was to develop a quick, flexible AI platform that can carry out a variety of cancer evaluation tasks, similar to ChatGPT.” “Our model proved to be very helpful in a variety of tasks pertaining to the detection, prognosis, and response to treatment of various cancers.”

CHIEF surpassed state-of-the-art AI techniques by up to 36.1% on all tasks when tested on over 19,400 photos from 32 separate datasets gathered worldwide. It also was better at distinguishing between patients with high and low survival rates and could accurately offer information about various tissue samples that were tested.

To increase CHIEF’s accuracy, the researchers intend to train it on pictures of non-cancerous illnesses, rare disorders, and pre-malignant tissues. To improve the model’s ability to detect cancer aggressiveness and forecast the results of innovative treatments, they also expect feeding it additional data.

An Expanding Role in Cancer Detection and Beyond. For some time now, scientists have been using AI to improve cancer and other disease detection, diagnosis, and therapy.

For instance, Cambridge researchers unveiled EMethylNET, an AI model that has a 98% success rate in identifying 13 different forms of cancer using DNA information from tissue samples. Trained on over 6,000 tissue samples, EMethylNET demonstrates how AI may detect cancer early by detecting DNA methylation, which is a key factor in the development of cancer.

An earlier model called CancerGPT (I’m not making that word up) predicted how drug combinations could affect cancer patients’ unusual tissues using a huge language model. It proved that when structured data and samples are limited, pre-trained models can be quite helpful. CancerGPT can provide important insights by generalizing predictions and using previous medical research, but researchers were still worried about possible AI hallucinations.

Google and iCAD collaborated to improve cancer screening with AI. Their AI-powered approach improved accessibility to life-saving breast cancer screenings and exceeded professional radiologists in accuracy, providing a workable solution in the face of a global radiology deficit.

Lastly, brain surgeons are using another AI tool called Sturgeon to help them diagnose malignancies in the central nervous system in real time with 90% accuracy.
​
CHIEF is open-source and may be downloaded from the project’s Github page, allowing researchers or anybody else to use it locally and add their own photos.

Fans will soon have a new option to follow player statistics from their seats during the game at the Los Angeles Rams’ state-of-the-art SoFi Stadium.

In a recent news release, the Rams announced a relationship with Genius Sports, a data, technology, and broadcast business that serves the media, sports, and betting industries.

Fans may now use Genius Sports’ data processing and artificial intelligence platform GeniusIQ to track various player metrics, like a player’s running speed and quarterback pass time, at every Rams home game.
The immersive highlights and data will be shown on SoFi Stadium’s Infinity Screen following their real-time analysis by AI.

The Rams stated in their press statement that “GeniusIQ applies powerful machine learning and AI to translate enormous volumes of Next Gen Stats data into a semantic understanding of football.” When the on-field action is at its most captivating, “it then powers real-time insights and brand activations, creating meaningful connections with the fans.”

 “Since opening SoFi Stadium in 2020, we continuously look to team up with best-in-class companies to set the gameday standard at the Rams House,” said Kathryn Kai-ling Frederick, Chief Marketing Officer, Los Angeles Rams. “We are thrilled to work with Genius Sports to deliver this unprecedented experience to our fans and grateful to our partner Verizon for elevating this content.”
​
To integrate their technology, Genius Sports has partnered with many other sports leagues worldwide, including NCAA, Liga MX, NASCAR, and the English Premier League.

August saw the Gates Foundation and the Foreign, Commonwealth and Development Office of the United Kingdom grant a $2.83 million grant to the British biotech business Legume Technology. AgFunderNews reports that the company intends to provide farmers in Africa with a cost-effective and safe way to enhance crop health and soil fertility with the additional cash.

Africa’s total soil fertility has declined over time, mostly because of excessive chemical fertilizer use that has sped up acidification.

Legume Technology has spent the last 20 years building bacterial and fungal microbial biofertilizerskeeping to fight that issue. These biofertilizers essentially create natural alternatives to chemical fertilizers by capturing nitrogen from the air and keeping it in the soil for the crops to absorb.

Redesigning the packaging as part of the company’s new strategy will enable more small-scale farmers to use biofertilizers in smaller containers.

Bruce Knight, a trained microbiologist who co-founded Legume Technology and serves as its managing director, said, “Within that small bag will be a microbe that has the power to transform the lives of millions of African smallholders by making their crops grow bigger and better, with more productive harvests, with no environmental side effects.”

Farmers have found it more and more difficult to sustain crop production in recent years. This is mostly because burning fossil fuels is causing air pollution, which is overheating our planet.

Certain places are no longer suitable for crops that are typically grown there because of the combination of changing climate conditions brought on by the overheating and unexpected extreme weather events.

Consequently, the development of innovative technologies that facilitate crop growth and food supply assurance for farmers while reducing their dependency on chemical fertilizers has become increasingly crucial. It has been shown that fertilizers have several downstream negative effects that eventually deteriorate ecosystem health.
​
“Biologicals are the future of agricultural productivity,” Knight stated.

Microsoft Copilot is a competitive alternative to ChatGPT since its inception, sometimes even surpassing OpenAI’s product. But Copilot trailed behind ChatGPT’s advancements—until now.

In a live event last week called “Microsoft 365 Copilot: Wave 2,” Microsoft revealed all the next improvements to its AI assistant, catering to both business and individual users.

The enhancements were striking, including cooperative methods to use AI as a team, group, or family, as well as further interaction with Microsoft 365 apps.

Copilot Pages. You should see a tab option to switch from Work to Web if you log into Copilot using your work or organization’s account. What Microsoft calls the “Work” tab is actually BizChat, a Copilot workflow that retrieves responses from your work data in Microsoft 365 apps. A new feature called Pages has been added to BizChat in an update.

Using Pages, you may update and add insights that Copilot generated for you based on your business data, then share them with your team so that you can work together. Pages are accessible via a shareable, editable link, much like any other type of editable document.

This section highlights multiplayer AI, a trend observed in recent releases from various AI businesses, including Salesforce’s Agentforce, a suite of collaborative and autonomous AI agents, and You.com’s collaborative AI assistants.

Customers of Microsoft 365 Copilot will be able to access the pages starting today, and in the upcoming weeks, free Microsoft Copilot users who are logged in with their Microsoft Entra account will access them as well.

Copilot in the Microsoft 365 apps. The ability of Microsoft Copilot to assist users within the Microsoft 365 apps—which have grown to be a reliable mainstay of many people’s workflows—is one of its greatest benefits. Microsoft is leveraging user feedback in Wave 2 to enhance user accessibility and expand its Copilot support within the applications.

To begin with, Copilot in Excel is widely accessible to all users and offers support for conditional formatting, data visualization, calculations, and more.

Microsoft also revealed the public preview of Copilot in Excel with Python, which enables users to interact with Python in Excel only through natural language. This implies that users don’t need to know how to write code to perform sophisticated analysis in Excel using Python, such as risk and forecasting analysis.

Microsoft released Narrative Builder for PowerPoint to all users. With the help of this function, you can quickly create an outline for your presentation by simply typing in a topic.

Now, Copilot in Teams can combine the information shared in the chat window with the contents of a real meeting to produce a summary of the entire meeting. For instance, everything from the chat will be included in the response if a user asks Copilot what they missed during the meeting. Next month, this feature will be made available to everyone.

Copilot in Outlook was too important to ignore: The tool that allows you to “Prioritize my inbox” now includes an analysis of your inbox and a flag for the most critical emails, along with a synopsis and the rationale for the flag. According to Microsoft, users will “soon” direct Copilot on what subjects or phrases to prioritize.

Microsoft states these capabilities will be made accessible to all users in late 2024, although they are not currently available.

Microsoft is releasing a new feature later this month that will enable users to reference web and work material quickly, including emails, PDFs, PowerPoints, and more, since users frequently accessed external resources when working in Microsoft Word. Another update for Copilot in Word is the “on-canvas start experience,” which lets users work together in real time while editing different areas of the text. These two Word features are currently accessible to most users.

Finally, Microsoft introduced Copilot in OneDrive, enabling users to search for items in their OneDrive repository, get summaries, and even perform file comparisons. OneDrive’s Copilot feature began trickling out to users last week and will soon be available to the public.

Copilot Agents. Microsoft unveiled Copilot agents, artificial intelligence (AI) assistants that can complete activities with as little or as much help from a user as needed. Microsoft, for instance, reports that while some agents are more basic “prompt-and-response” agents, others can operate entirely on their own.
Using the new agent builder, which is powered by Copilot Studio and enables users to create agents in Sharepoint or BizChat, users may create Copilot agents. Then, just like any other teammate, these agents can be called upon via the “@” symbol in the 365 applications.
​
The agent builder in BizChat and Copilot agents are accessible right now and will be made available to all clients in the upcoming weeks. Microsoft has announced that the public preview of Copilot agents and the agent builder in SharePoint will begin in early October.

Thomson Reuters, a global content and technology company focused on the public accounting profession, announced last week the launch of Thomson Reuters Audit Intelligence. This suite of Al-powered data-driven audit solutions is designed to transform audit practices by reducing errors and enhancing confidence, while enabling auditors to realize efficiency gains.

The first tool to debut in the suite, named Audit Intelligence Analyze, employs Al and machine learning to simultaneously enhance both audit quality and efficiency. The solution focuses on high-risk areas by efficiently segmenting audit testing populations based on risk level and reducing the overall number of items to be tested. With advanced anomaly detection, Analyze integrates directly within existing workflows, identifying unusual items often overlooked by humans and automatically generating all necessary audit documentation.

Dave Wyle, general manager of Audit at Thomson Reuters, emphasized the importance of using technology to modernize audit practices to stay competitive. "Today's announcement underscores our commitment to investing in and developing products that enable data-driven, Al-powered audits. We aim to provide these solutions swiftly to our customers, maximizing the benefits of Al in their audits, while supporting firms to implement solutions at a pace that works for their businesses," he said. "Our new solution, Analyze, can cut sample sizes by half, saving valuable time for auditors and their clients. This makes audits more efficient and allows auditors to focus on higher-risk areas, improving overall audit quality."

The Audit Intelligence suite will integrate CoCounsel, Thomson Reuters professional-grade GenAl assistant, from next year, as part of the company's broader vision to provide a GenAl assistant to every professional it serves. Auditors will be able to leverage CoCounsel alongside Thomson Reuters Guided Assurance to automate the completion of audit program steps and checklists.

Future additions to the suite will include Audit Intelligence Test, which automates substantive testing and verification by dynamically tracing accounting transactions to banking activity and supporting documentation, and Audit Intelligence Plan, which merges full data populations with technology-driven analytics to empower auditors' planning and risk assessment processes.

Key benefits of Thomson Reuters Audit Intelligence Analyze
• Increased efficiency: Automation of manual processes allows for smaller, risk-focused samples. This enables auditors to concentrate on high-risk areas, enhancing efficiency and reducing overall firm risk while improving audit quality. Time savings of 30 minutes to two hours are achieved in both client data provision and auditor sample selection and documentation.

• Enhanced accuracy: Automated analysis using Al and machine learning increases confidence in audit results, integrating seamlessly into existing audit processes without requiring new methodologies.

• Improved transparency: The solution is fully configurable, providing a clear and transparent process for selecting and understanding advanced risk identification techniques.
​
Available now as part of an early adopter program, Audit Intelligence Analyze is expected to debut in the United States in the fall of 2024, followed by the UK in 2025. The solution is integrated into both Thomson Reuters Cloud Audit Suite and directly into audit methodologies and workflows. It can also be used as a standalone by firms with their own methodologies.

It is simple to have your computer or tablet screen visually hacked in public, which allows someone to steal private information. Plastic screen coverings are one technique to stop this from happening, as they make it harder for someone to view the document unless they are directly in front of it. Microsoft might develop a technology that encrypts text graphically as you read it.

A system that encrypts a document so that only the reader can decipher it is described in a recent Microsoft patent. By changing the letters in the text section of the original document, the approach encrypts the document. The document is then shown in a private mode using an alpha-blended version that combines the encoded and original copies of the text. Next, the system takes advantage of eye tracking to show the user the decoded portion of the document at their fixation point and the encoded version at the periphery of their field of vision.

This issue is currently addressed by a wide range of products on the market, from hardware privacy screens to software that uses webcams and eye tracking to let light from the screen enter the eyes of the intended user while blocking light from the screen from entering the eyes of onlookers. They are all limited, though.
Because it is challenging to attain 100% light efficiency, privacy screens make the display appear dull to the intended user’s eyes. In the meantime, reading the text becomes more challenging when non-fixed portions of the screen are arbitrarily blurred or dimmed—it becomes like reading through a telescope.

The Ponemon Institute claims that visual hacking is a common issue. The institute carried out an experiment in the US in 2015 that showed how simple it was to use visual hacking to get confidential firm information. In 2016, 3M supported an enlarged version of this project with offices in China, France, Germany, India, Japan, South Korea, and the UK.
​
According to the 3M investigation, in 91% of the worldwide trials, information was visually hacked by an undercover white hat hacker. An undercover visual hacker was able to collect information in 15 minutes or less in nearly half of the global trials. Only 32% of the hacker efforts were successful, and 52% of the sensitive data was taken from staff computer screens. The outcomes matched those of the smaller study conducted by the Ponemon Institute.

According to Citigroup Inc., more positions in the banking sector will probably be replaced by artificial intelligence (AI) than in any other. Technology is going to transform consumer finance. It will also make employees more productive.

According to Citigroup’s most recent AI assessment, which was published last week, roughly 54% of banking professions might be automated. AI technology has the potential to enhance an additional 12% of jobs within the business.

The biggest banks in the world have been using artificial intelligence (AI) more frequently over the past year because of its ability to boost output and cut expenses. According to a recent analysis by Citigroup, artificial intelligence (AI) might boost the banking sector by $170 billion by 2028.

Citigroup intends to let its 40,000 programmers experiment with different AI tools. The business has already reviewed hundreds of pages of regulatory submissions swiftly using generative AI, which can generate content from basic prompts.

“We’re currently focused on bringing it from the lab to the manufacturing floor,” CEO of Citigroup Jane Fraser stated during the digital money symposium held by the organization. She continued by saying that the bank is investigating the use of AI to improve cybersecurity and offer affluent clients individualized investing advice.

CEO Jamie Dimon of JPMorgan Chase & Co. thinks AI might reduce the workweek to just 3.5 days, which is why the bank is employing AI specialists. While ING Group uses AI to find possible defaulters, Deutsche Bank AG uses it to evaluate the portfolios of wealthy clients.

According to David Griffiths, Chief Technology Officer of Citigroup, generative AI “has the potential to revolutionize the banking industry and improve profitability,” in a statement released alongside the new findings. “At Citi, our goal is to leverage AI in a responsible and safe manner to increase the power of Citi and our people.”
​
According to Citigroup, AI may not cause a decrease in the total number of workers in the banking sector, even if it replaces some occupations. It is likely that financial institutions will need to employ a large number of AI managers and compliance officers to make sure that rules are observed when using AI technology.

AI PCs’ increased power and efficiency are poised to change computing. Is the computer you’re considering an AI PC, though? We’ll examine their special characteristics and whether it makes sense to upgrade to these cutting-edge devices.

How Does It Work? An AI PC is a powerful device built specifically to tackle artificial intelligence (AI) tasks with ease. It’s not your typical computer. The device processes everything in-house, doing away with the requirement for cloud data processing.

How then does this all operate? AI PCs, in contrast to conventional computers, have specialized hardware, such as GPUs (Graphics Processing Units) and AI accelerators, which improve their machine-learning capabilities.

These computers include sophisticated software algorithms that effectively manage data-intensive artificial intelligence activities, allowing for quicker processing and real-time analytics. AI PCs can accomplish activities like complex data analysis and natural language processing far more accurately than traditional computers, thanks to these potent features.

When Will AI Computers Be Available? AI personal computers are already beginning to be released; big businesses such as AMD and Intel are leading the push. With the release of its Intel Core Ultra portfolio in December 2023, Intel has advanced significantly. Integrated NPUs (Neural Processing Units) in this range improve laptop AI tasks.

By 2025, Intel hopes to have sold 100 million AI PCs, indicating the industry’s rapid expansion and uptake. However, you are not restricted to using only Intel as an option. AMD and Qualcomm are adding AI capabilities to their CPUs, so during 2024, additional AI PC releases should be available from them.

How Do You Easily Identify an AI PC? its software and hardware specs define an AI PC. Look for references to NPUs, AI-specific accelerators, sophisticated GPUs, and higher-than-average RAM and storage specifications for hardware. These characteristics are essential for managing complicated algorithms and big datasets. However, the new Copilot key on the keyboard is the easiest method to identify a Windows computer as an AI PC.

AI PCs frequently have pre-installed software that makes use of AI to enhance functionality. Better voice assistants, cutting-edge security features like facial recognition, and applications that customize information based on user behavior are examples of this.

AI PCs perform noticeably faster and more efficiently when handling massive amounts of data, executing complex simulations, and producing high-quality visuals.

Is Buying an AI Computer Better Than a Normal One? Whether to switch from a conventional PC to an AI PC will mostly depend on your individual requirements. Consider the following benefits and drawbacks to aid in your decision.

Advantages:

• Enhanced Performance and Efficiency: AI PCs are perfect for tasks involving AI and data analytics since they can process and analyze data at previously unheard-of speeds.

• Personalized Experiences: AI-powered algorithms enable these PCs to anticipate user needs based on usage patterns, provide personalized recommendations, and tailor interfaces to suit individual tastes.

• Enhanced Security: Better threat detection and real-time security measures are made possible by AI PCs’ inherent capabilities.

• Compatibility With Emerging Technologies: AI PCs are designed to be easily integrated with new technologies, so they will continue to be useful even as new advancements take place.

Disadvantages:

• Higher Initial Cost: Because AI PCs have sophisticated technology and capabilities, they come with a higher price tag.

• Complexity and Learning Curve: Users who are not tech-savvy may find the advanced technologies built into AI PCs daunting.

• Privacy Issues: AI PCs’ sophisticated data-processing powers give rise to worries about data privacy and possible abuse.

• Energy Usage for High-Performance Tasks: Artificial intelligence tasks require a lot of resources. Therefore, if you run them on AI PCs instead of ordinary PCs for routine tasks, you may end up using significantly more energy.
​
AI PCs undoubtedly mark a major advancement in computer technology, and AI is undoubtedly what Windows will look like in the future. These devices may be appropriate if you require high-performance machines that can handle intricate datasets and carry out tasks connected to artificial intelligence. While the price and complexity may put off some users, AI PCs may have revolutionary effects for individuals working in software development and data research, among other professions.

Many people focus on the potential benefits of generative AI for kids when thinking about its application in the classroom, but teachers can gain just as much, if not more, from the technology. A free Google Generative AI Educators course was launched last week by Google and MIT Responsible AI for Social Empowerment and Education (RAISE) to assist middle and high school educators in utilizing generative AI tools to improve their workflow and the learning environment for their students.

According to the press release, the self-paced, two-hour course teaches educators how to leverage generative AI to save time on routine chores like email composing, adapting content for various reading levels, creating innovative evaluations, organizing activities around students' interests, and more. By feeding their current lesson plan into the generative AI models, educators can even learn how to use generative AI to assist with one of the most time-consuming tasks: lesson planning. This will provide ideas for what to do next in the classroom.

"This course empowers educators to confidently integrate AI into their teaching, creating richer and more accessible learning experiences for all students," MIT RAISE Director Cynthia Breazeal said.

Teachers and administrators can access the course's five modules, each lasting no more than forty minutes, on the Generative AI for Educators webpage.

According to the press release, several school districts across the nation also intend to offer the course, including the Anaheim Union High School District in California, the Miami-Dade County Public Schools in Florida, the Chicago Public Schools in Illinois, the Albuquerque Public Schools in New Mexico, the Glenpool Public Schools in Oklahoma, and the Sun Prairie Area School District in Wisconsin.

"In a rapidly evolving world, our teachers cannot afford to fall behind in accessing powerful generative tools that will help them develop new approaches to teaching and learning," Michael Matsuda, the superintendent at Anaheim Union High School District, said.
​
The company that created ChatGPT, OpenAI, previously emphasized the advantages of providing teachers with shared use cases and generative AI tools, as well as examples of how teachers throughout the nation are already utilizing the technology. Examples given by OpenAI included teaching students about critical thinking, creating instructional materials, helping non-native English speakers with their English, and employing a chatbot to role-play discussions intended for students.