Category: ai

The first of these modules involves placing “structural anchors” throughout the spreadsheet to help the LLM understand what’s going on better. It then removes “distant, homogeneous rows and columns” to produce a condensed “skeleton” version of the table. Index translation addresses the challenge caused by spreadsheets with numerous empty cells and repetitive values, which use up too many tokens. “To improve efficiency, we depart from traditional row-by-row and column-by-column serialization and employ a lossless inverted index translation in JSON format,” Microsoft wrote. “This method creates a dictionary that indexes non-empty cell texts and merges addresses with identical text, optimizing token usage while preserving data integrity.” […]

After conducting a “comprehensive evaluation of our method on a variety of LLMs” Microsoft found that SheetCompressor significantly reduces token usage for spreadsheet encoding by 96%. Moreover, SpreadsheetLLM shows “exceptional performance in spreadsheet table detection,” which is the “foundational task of spreadsheet understanding.” The new LLM builds on the Chain of Thought methodology to introduce a framework called “Chain of Spreadsheet” (CoS), which can “decompose” spreadsheet reasoning into a table detection-match-reasoning pipeline.

In March, after analysing 22,000 tasks in the UK economy, covering every type of job, a model created by the Institute for Public Policy Research predicted that 59% of tasks currently done by humans — particularly women and young people — could be affected by AI in the next three to five years. In the worst-case scenario, this would trigger a “jobs apocalypse” where eight million people lose their jobs in the UK alone…. Darrell West, author of The Future of Work: AI, Robots and Automation, says that just as policy innovations were needed in Thomas Paine’s time to help people transition from an agrarian to an industrial economy, they are needed today, as we transition to an AI economy. “There’s a risk that AI is going to take a lot of jobs,” he says. “A basic income could help navigate that situation.”

AI’s impact will be far-reaching, he predicts, affecting blue- and white-collar jobs. “It’s not just going to be entry-level people who are affected. And so we need to think about what this means for the economy, what it means for society as a whole. What are people going to do if robots and AI take a lot of the jobs?”

Nell Watson, a futurist who focuses on AI ethics, has a more pessimistic view. She believes we are witnessing the dawn of an age of “AI companies”: corporate environments where very few — if any — humans are employed at all. Instead, at these companies, lots of different AI sub-personalities will work independently on different tasks, occasionally hiring humans for “bits and pieces of work”. These AI companies have the potential to be “enormously more efficient than human businesses”, driving almost everyone else out of business, “apart from a small selection of traditional old businesses that somehow stick in there because their traditional methods are appreciated”… As a result, she thinks it could be AI companies, not governments, that end up paying people a basic income.

AI companies, meanwhile, will have no salaries to pay. “Because there are no human beings in the loop, the profits and dividends of this company could be given to the needy. This could be a way of generating support income in a way that doesn’t need the state welfare. It’s fully compatible with capitalism. It’s just that the AI is doing it.”

The document describes a project that uses Strawberry models with the aim of enabling the company’s AI to not just generate answers to queries but to plan ahead enough to navigate the internet autonomously and reliably to perform what OpenAI terms “deep research,” according to the source. This is something that has eluded AI models to date, according to interviews with more than a dozen AI researchers. Asked about Strawberry and the details reported in this story, an OpenAI company spokesperson said in a statement: “We want our AI models to see and understand the world more like we do. Continuous research into new AI capabilities is a common practice in the industry, with a shared belief that these systems will improve in reasoning over time.”

On Tuesday at an internal all-hands meeting, OpenAI showed a demo of a research project that it claimed had new human-like reasoning skills, according to Bloomberg, opens new tab. An OpenAI spokesperson confirmed the meeting but declined to give details of the contents. Reuters could not determine if the project demonstrated was Strawberry. OpenAI hopes the innovation will improve its AI models’ reasoning capabilities dramatically, the person familiar with it said, adding that Strawberry involves a specialized way of processing an AI model after it has been pre-trained on very large datasets. Researchers Reuters interviewed say that reasoning is key to AI achieving human or super-human-level intelligence.

A study published in the June issue of IEEE Transactions on Software Engineering evaluated the code produced by OpenAI’s ChatGPT in terms of functionality, complexity and security. The results show that ChatGPT has an extremely broad range of success when it comes to producing functional code — with a success rate ranging from anywhere as poor as 0.66 percent and as good as 89 percent — depending on the difficulty of the task, the programming language, and a number of other factors. While in some cases the AI generator could produce better code than humans, the analysis also reveals some security concerns with AI-generated code.

The study tested GPT-3.5 on 728 coding problems from the LeetCode testing platform — and in five programming languages: C, C++, Java, JavaScript, and Python. The results?

Overall, ChatGPT was fairly good at solving problems in the different coding languages — but especially when attempting to solve coding problems that existed on LeetCode before 2021. For instance, it was able to produce functional code for easy, medium, and hard problems with success rates of about 89, 71, and 40 percent, respectively. “However, when it comes to the algorithm problems after 2021, ChatGPT’s ability to generate functionally correct code is affected. It sometimes fails to understand the meaning of questions, even for easy level problems,” said Yutian Tang, a lecturer at the University of Glasgow. For example, ChatGPT’s ability to produce functional code for “easy” coding problems dropped from 89 percent to 52 percent after 2021. And its ability to generate functional code for “hard” problems dropped from 40 percent to 0.66 percent after this time as well…

The researchers also explored the ability of ChatGPT to fix its own coding errors after receiving feedback from LeetCode. They randomly selected 50 coding scenarios where ChatGPT initially generated incorrect coding, either because it didn’t understand the content or problem at hand. While ChatGPT was good at fixing compiling errors, it generally was not good at correcting its own mistakes… The researchers also found that ChatGPT-generated code did have a fair amount of vulnerabilities, such as a missing null test, but many of these were easily fixable.

“Interestingly, ChatGPT is able to generate code with smaller runtime and memory overheads than at least 50 percent of human solutions to the same LeetCode problems…”

But the decision regarding Meta will “very likely” encourage other companies to refrain from being transparent in the use of data in the future, said Ronaldo Lemos, of the Institute of Technology and Society of Rio de Janeiro, a think-tank. “Meta was severely punished for being the only one among the Big Tech companies to clearly and in advance notify in its privacy policy that it would use data from its platforms to train artificial intelligence,” he said. Compliance must be demonstrated by the company within five working days from the notification of the decision, and the agency established a daily fine of 50,000 reais ($8,820) for failure to do so. In a statement, Meta said the company is “disappointed” by the decision and insists its method “complies with privacy laws and regulations in Brazil.”

“This is a step backwards for innovation, competition in AI development and further delays bringing the benefits of AI to people in Brazil,” a spokesperson for the company added.

The very-high-level, harder-than-it-sounds solution Anthropic is proposing is creating challenging benchmarks with a focus on AI security and societal implications via new tools, infrastructure and methods.

Building products using Machine Learning and data is still difficult. The tooling infrastructure is still very immature and the non-standard combination of data and software creates unforeseen challenges for engineering teams. But in my views, a lot of the failures come from this explosive cocktail of ritualistic Machine Learning:

– Weak software engineering knowledge and practices compounded by the tools themselves;
– Knowledge gap in mathematical, statistical, and computational methods, encouraged black boxing API;
– Ill-defined range of competence for the role of data scientist, reinforced by a pool of candidates with an unusually wide range of backgrounds;
– A tendency to follow the hype rather than the science. –

What can you do?

– Hold your data scientists accountable using Science.
– At a minimum, any AI/ML project should include an Exploratory Data Analysis, whose results directly support the design choices for feature engineering and model selection.
– Data scientists should be encouraged to think outside-of-the box of ML, which is a very small box
– Data scientists should be trained to use eXplainable AI methods to provide context about the algorithm’s performance beyond the traditional performance metrics like accuracy, FPR, or FNR.
– Data scientists should be held at similar standards than other software engineering specialties, with code review, code documentation, and architectural designs.

The article concludes, “Until such practices are established as the norm, I’ll remain skeptical of Data Science.”

ELIZA, a system pre-programmed with responses but with no large language model (LLM) or neural network architecture, was judged to be human just 22% of the time. GPT-3.5 scored 50% while the human participant scored 67%. “Machines can confabulate, mashing together plausible ex-post-facto justifications for things, as humans do,” Nell Watson, an AI researcher at the Institute of Electrical and Electronics Engineers (IEEE), told Live Science. “They can be subject to cognitive biases, bamboozled and manipulated, and are becoming increasingly deceptive. All these elements mean human-like foibles and quirks are being expressed in AI systems, which makes them more human-like than previous approaches that had little more than a list of canned responses.” Further reading: 1960s Chatbot ELIZA Beat OpenAI’s GPT-3.5 In a Recent Turing Test Study

The NPU allows the Raspberry Pi 5 to perform AI tasks such as object and facial recognition, human pose analysis, and more. Using an NPU frees up the Raspberry Pi 5’s CPU, allowing it to focus on other tasks, making your projects more efficient and powerful. The Raspberry Pi AI kit is also compatible with the Raspberry Pi Active Cooler, ensuring optimal performance without overheating. Additionally, you can purchase a clear protective layer to prevent damage to the board, giving you peace of mind while working on your projects. The AI kit is priced at $70. It’s available from Raspberry Pi Approved Resellers, including PiHut, PiShop.us, and CanaKit.