AI driven Google searches could require as much electricity as it takes to power a country the size of Ireland, Europe is demanding Elon Musk and Mark Zuckerberg take immediate action against disinformation and a Polish startup features the technology that Google thought was too dangerous to release.
These and more top tech stories on Hashtag Trending.
I’m your host Jim Love, CIO of IT World Canada and Tech News Day in the US.
The surge in interest and application of large language models (LLMs) and generative AI has sparked concerns about a potential spike in datacenter electricity consumption, according to a paper by Alex de Vries, a researcher at a university in Amsterdam. The paper highlights that while the training phase of AI models is often scrutinized for its energy use, the inference phase, or operational use of the trained model, might also significantly contribute to an AI model’s life-cycle costs.
For instance, to support ChatGPT, OpenAI required 3,617 servers with a total of 28,936 GPUs, implying an energy demand of 564 MWh per day. If we got the math right, it would power about 20,000 US households.
But even more disturbing is a quote from Alphabet’s chairman that if every search became a Large Language Model transaction, it would “likely cost 10 times more than a standard keyword search.”
The researcher extrapolated from this that the electricity needed could equate to Ireland’s annual consumption of 29.3 TWh.
However, the paper also acknowledges that this represents a worst-case scenario, assuming full-scale AI adoption with current technology, which is “unlikely to happen rapidly.” But the paper also notes that what is referred to as the Jevons paradox might apply – where increased efficiency stimulates demand, potentially wiping out any savings from that new technology.
Sources: The Register
The European Union (EU) has issued stern warnings to both Elon Musk’s X and Mark Zuckerberg’s Meta about the proliferation of misinformation and “violent and terrorist” content, particularly in the context of the Israel-Hamas conflict.
The EU, through its industry chief Thierry Breton, has mandated that these platforms demonstrate “timely, diligent, and objective action” in countering the spread of disinformation and comply with European law, under the Digital Services Act and Terrorist Content Online Regulation, which requires monitoring and removal of illegal content.
The platforms have been given a 24-hour window to respond and detail the measures taken. This comes amidst a surge of doctored images, mislabeled videos, and misleading content circulating on these platforms, sowing confusion and tension amidst the ongoing conflict.
Some agencies are not waiting for action to be taken. The Federal Anti-Discrimination Agency (FADA) of Germany has pulled out of X, citing a significant increase in hate speech and various forms of hostility since Musk took ownership.
Mozilla, the entity behind the Firefox browser, has moved to defend against another form of disinformation – namely fake reviews. It has integrated a “fake reviews detector” into its platform, following its acquisition of Fakespot in May. Fakespot, a startup specializing in identifying fake reviews and news through its website and browser extension, has been utilized to detect fraudulent reviews on platforms like Amazon, Yelp, TripAdvisor, Walmart, and eBay, employing an A-to-F grading scale.
The Review Checker feature is slated for release on Firefox version 120 for desktop and Android on November 21, 2023.
Fakespot utilizes advanced AI and ML systems to identify patterns and similarities among reviews, flagging those likely to be deceptive and potentially hindering efforts to artificially boost product rankings through fake reviews using AI technologies like ChatGPT.
PimEyes, a website developed by a Polish startup, offers facial recognition tools available to the public. These tools allow users to upload a photo of a person’s face and, using AI, scans the internet to find images of that person, even those they might not be aware exist. There is a free version of the software and a paid version that alerts users when a new photo appears online.
While it claims to help people monitor their online presence, it has sparked controversy due to its potential use as a surveillance tool for stalkers, its collection of images of minors, and adding pictures of deceased individuals to its database without consent.
PimEyes CEO, emphasizes that the tool does not identify people but identifies websites that feature images similar to the search material. However, privacy advocates and experts express concerns about the potential misuse of such technology, especially in the absence of federal laws governing facial recognition technology in the U.S.
Despite their disclaimers, PimEyes now blocks access in 27 countries, including Iran, China and Russia, over fears government authorities could use the service to target protesters and dissidents.
A journalist with the Hill Times quotes Eric Schmidt as far back as 2011, as saying “this was the one technology that Google had developed and decided to hold back, that it was too dangerous in the wrong hands.”
That’s the top tech news stories for today. For more fast reads on top stories, check us out at TechNewsDay.com or ITWorldCanada.com on the homepage.
Hashtag Trending goes to air 5 days a week with a special weekend interview show we call “the Weekend Edition.”
You can get us anywhere you get audio podcasts and there is a copy of the show notes at itworldcanada.com/podcasts
I’m your host, Jim Love – have a Thrilling Thursday.