IRS using AI to catch rich people and tax-dodging corps

AI in brief The US Internal Revenue Service has said it will use AI software to go after wealthy individuals and corporations violating tax laws.

The tax agency said experts in data science and tax enforcement have been applying "cutting-edge machine learning technology" to find anomalies in partnership taxes, general income taxes and accounting, and international taxes. It's not clear what type of AI algorithms or models are being used exactly. 

These audits will be mostly run on taxpayers with incomes over $1 million that owe more than $250,000 in taxes and might be avoiding paying their dues. The IRS also said it has found "ongoing discrepancies" on balance sheets involving partnerships with over $10 million in assets, a sign of "potential non-compliance."

It warned that by the end of the month, it will launch investigations of 75 of the largest partnerships in the US, a group made up of hedge funds, real estate investment partnerships, publicly traded partnerships, large law firms and others.

"The nation relies on the IRS to collect funding for every critical government mission - from keeping our skies safe, our food safe and our homeland safe," IRS Commissioner Danny Werfel, said in a statement. "It's critical that the agency addresses fundamental gaps in tax compliance that have grown during the last decade."

"We will increase our compliance efforts on those posing the greatest risk to our nation's tax system, whether it's the wealthy looking to dodge paying their fair share or promoters aggressively peddling abusive schemes. These steps are critical for the future of the nation's tax system," he said.

AI could be a bigger shift in our lives than the internet, says Google CEO

Google CEO Sundar Pichai believes AI will be the biggest technology shift in our lifetimes, and that its impact could be even bigger than the creation of the internet itself.

As Google celebrates its 25th birthday this week, Pichai reflected on the company's vision over time. The company has arguably had the largest influence on the internet, and has shaped the way netizens use the technology. 

Pichai said search is still Google's biggest moonshot, and that the most important tool "to develop services that improve the lives of as many people as possible – to do things that matter" is AI.

"Over time, AI will be the biggest technological shift we see in our lifetimes," he said. "It's bigger than the shift from desktop computing to mobile, and it may be bigger than the internet itself. It's a fundamental rewiring of technology and an incredible accelerant of human ingenuity."

Google ramped up its efforts to deploy generative AI features into its products. It rolled out its search engine chatbot Bard to compete with Microsoft's Bing AI, and injected new tools to help users automatically produce text in its Google Workspace apps. It's also reportedly preparing to release its most powerful model yet codenamed Gemini. 

"Making AI more helpful for everyone, and deploying it responsibly, is the most important way we'll deliver on our mission for the next 10 years and beyond," Pichai said.

Anthropic releases Claude Pro

AI safety lab and large language model-maker startup, Anthropic, launched the paid version of its latest Claude 2 chatbot this week for users in the UK and US. 

Claude 2 was released about three months shortly after the older Claude bot was announced. It has better coding, math, and reasoning skills compared to its predecessor, and a very long 100,000-token context window. Tokens are short snippets of words, and having a larger context window means more text from documents, articles, or reports can be included in the input prompt.

The model can then better perform natural language tasks on more data, like summarisation and classification, and generate more text too. Claude Pro supports 5X more usage than the limits placed on using the free web app. The paid version will cost users $20 or £18 per month in the US and UK, respectively. 

Customers will get priority access to the chatbot during periods of high-traffic compared to non-paying users, and will get to try new features first. 

"This means you can level up your productivity across a range of tasks, including summarizing research papers, querying contracts, and iterating further on coding projects — like this recent demo of building an interactive map," the upstart said. 

GPT and LLM are now official words on Dictionary.com

Common acronyms and lingo often used in AI, like GPT and LLM, are official words in Dictionary.com.

LLM stands for large language model, of course, whilst GPT is short for generative pre-trained transformer, the computational architecture powering the latest AI chatbots. Those aren't the only latest additions on the website, however.

• TSMC warns AI chip crunch will last another 18 months
• NixCon drops Palmer Luckey's AI combat drone maker Anduril as sponsor due to military ties
• Microsoft to shield paid-up Copilot customers from any AI copyright brawls it starts
• AI to replace 2.4 million jobs in the US by 2030, many fewer than other forms of automation

The Verge noted that dictionary.com even updated its entry for hallucinate to describe the term used when generative AI models veer off track and produce text laden with inaccuracies and errors.

Hallucinate now also means a machine learning program producing "false information contrary to the intent of the user and [presenting] it as if true and factual." The website added generative AI as a noun too, describing it as "artificial intelligence that is designed to process prompts from users and respond with text, images, audio, or other output that is modeled on a training data set." ®