AI-driven Google searches could require as much electricity as it would take to power a country the size of Ireland, and Europe has seen Elon Musk and Mark Zuckerberg report disinformation The Polish startup has technology that Google considered too dangerous to release publicly.
More top technology stories on Hashtag Trends.
I’m your host, Jim Love. He is the CIO of IT World in Canada and Tech News Day in the US.
The surge in interest and applications in large-scale language models (LLMs) and generative AI has raised concerns about the potential for data center power consumption to skyrocket. According to a paper by Alex de Vries, a researcher at the University of Amsterdam.. The paper argues that while energy usage is often scrutinized during the training phase of an AI model, the inference phase, or operational use of the trained model, can also contribute significantly to the lifecycle cost of an AI model. I am emphasizing one thing.
For example, to support ChatGPT, OpenAI requires 3,617 servers with a total of 28,936 GPUs, implying an energy demand of 564 MWh per day. If the math is correct, it could power about 20,000 homes in the United States.
But even more alarming is what Alphabet’s chairman says: What if every search becomes a transaction in a large language model? “It can cost 10 times more than a standard keyword search.”
From this, the researchers estimated that the electricity required would be equivalent to Ireland’s annual consumption of 29.3 TWh.
However, the paper also acknowledged that with current technology, this is a worst-case scenario and is “unlikely to occur quickly” assuming full-scale adoption of AI. But the paper also points out that something called the Jevons paradox, where efficiency gains can stimulate demand and wipe out any savings from new technology, could apply.
source: register
The European Union has issued stern warnings to both Elon Musk’s Mr. X and Mark Zuckerberg’s Mr. Meta about the spread of misinformation and “violent and terrorist” content. Especially in the context of the conflict between Israel and Hamas.
The EU, through industry chief Thierry Breton, has mandated that these platforms demonstrate “timely, diligent and objective action” to counter the spread of disinformation and comply with European law. under the Digital Services Act and Online Terrorist Content; Regulations requiring monitoring and removal of illegal content.
The platform has a 24-hour window We will respond and explain in detail the measures taken. This comes as doctored images, mislabeled videos and misleading content proliferate on these platforms, creating confusion and tension amid the ongoing conflict.
Some government agencies are not waiting for action to be taken. Germany’s Federal Anti-Discrimination Agency (FADA) withdrew from Company X, citing a significant increase in hate speech and various forms of hostility since Musk took over.
source: D.W., CNBC, BBC
Mozilla, the force behind the Firefox browser, is working to defend against another form of disinformation: fake reviews. After acquiring Fakespot in May, the company integrated a “fake review detector” into its platform. Fakespot is a startup specializing in identifying fake reviews and news through websites and browser extensions used to detect fraudulent reviews on platforms such as Amazon, Yelp, TripAdvisor, Walmart, and eBay. and uses a rating scale from A to F.
The review checker feature is scheduled to be released in Firefox version 120 for desktop and Android on November 21, 2023.
Fakespot leverages advanced AI and ML systems to identify patterns and similarities between reviews. Through fake reviews using AI technology like ChatGPT, we flag those that are potentially deceptive and can hinder efforts to artificially boost product rankings.
source: mashable
PimEyes is a website developed by a Polish startup that provides publicly available facial recognition tools. These tools allow users to upload a photo of a person’s face and use AI to scan the internet for images of that person that you may not have known existed.The software has a free version There is also a paid version that alerts users when new photos appear online.
It claims to help people monitor their online presence, but it also warns that it can be used as a surveillance tool for stalkers, collect images of minors, and kill people without their consent. This has caused controversy by adding photos of individuals who have been arrested to the database.
PimEyes CEO emphasizes that the tool does not identify individuals, but websites. We feature images similar to the search material. But privacy advocates and experts have raised concerns about the potential for such technology to be misused. Especially since there is no federal law governing facial recognition technology in the United States.
Despite the disclaimer, PimEyes currently blocks access in 27 countries. Iran, China and Russia are concerned that government authorities could use the service to target protesters and dissidents.
Quoted by Hill Times journalist Eric Schmidt In 2011, he said, “This was one of the technologies that Google had developed, but we decided not to develop it because it was too dangerous in the wrong hands.”
source: NPR
That’s today’s top technology news. To read top stories faster, visit our homepages at TechNewsDay.com or ITWorldCanada.com.
Hashtag Trends will air a special weekend interview show called “Weekend Edition” five days a week.
You can reach us wherever you get audio podcasts. A copy of the show notes can be found at itworldcanada.com/podcasts.
I’m your host, Jim Love – have a thrilling Thursday.
