New GPT-4 has more misinformation than predecessors? TikTok clamps down on deepfakes and AWS moves from diesel to vegetable oil.
I’m your host Jim Love, CIO of IT World Canada and TechNewsDay in the US – here’s today’s top tech news stories.
Last week, OpenAI debuted its newest generative AI tool, GPT-4 which was unveiled as a quantum leap from its predecessor GPT-3.5. It is supposedly
smarter, more creative, and safer version of its AI technology” Open AI’s site noted that “GPT-4 is 82 per cent less likely to respond to requests for disallowed content and 40 per cent more likely to produce factual responses than GPT-3.5.”
But according to a report from NewsGuard, GPT-4 may have “scored in the 90th percentile on the bar exam, but the latest version of OpenAI’s artificial intelligence software scored zero percent in an exercise assessing its ability to avoid spreading significant misinformation.”
NewsGuard is a service that uses trained journalists to rate news and information sites. It considers itself an independent, third-party, despite the fact that, according to an Axios report, it is backed by Microsoft, who also have invested heavily in OpenAI.
NewsGuard used the same test to observe how both GPT-3.5 and 4 would respond to a series of leading prompts relating to 100 false narratives, including the Sandy Hook elementary school shooting and COVID-19 vaccines.
GPT 3.5 generated 80 of the 100 false narratives and GPT-4 responded with misleading claims for all of the false narratives. Newsguard noted “ChatGPT-4, is actually more susceptible to generating misinformation — and more convincing in its ability to do so — than its predecessor.”
The report notes AI companies can face greater misinformation problems as their technology gets more sophisticated at delivering answers that look authoritative.
Source: Axios
In yet another in a stream of AI announcements, Microsoft said yesterday that it will allow users to generate images through its new AI-enabled Bing.
This new capability, dubbed “Bing Image Creator” will be powered by DALL-E, OpenAI’s generative image generator.
It’s now rolling out to users in the Bing preview and will only be available through Bing’s Creative Mode. It’ll come to Bing’s Balanced and Precise modes in the future. The new image generator will also be available in the Edge sidebar.
Microsoft also detailed the guardrails on the new feature saying, “we have put controls in place that aim to limit the generation of harmful or unsafe images. When our system detects that a potentially harmful image could be generated by a prompt, it blocks the prompt and warns the user.”
Source: TechCrunch
Meanwhile, Google is either late to the party or proving that slow but steady might win the AI race. After a false start last month, marred by an error in an answer that turned into an internet sensation, Google is now allowing some users in the U.S. and U.K. to start accessing the experimental version of Bard, that it previewed a month ago.
Bard is powered by a smaller, optimized version of Google’s LaMDA large language model.
Calling it a “phased in” approach, the company said that Bard will be updated with newer, more capable models over time.
The new Bard will offer three options, or “drafts” of its answer to a question and, to allow users to check the accuracy, it offers what a lot of us use to be certain of an answer – it allows you a button so you can “Google it.”
Source: Axios
And in even more AI news, Adobe is the latest to introduce its generative AI solution set, with the release of its new family of AI models called Firefly.
Firefly is an expansion of the generative AI tools Adobe introduced in Photoshop, Express and Lightroom during its annual Max conference last year. It lets users create and edit objects, composites and effects by simply describing them.
Adobe VP of generative AI, Alexandru Costin told TechCrunch in an email interview, “Firefly is the next step on our AI journey — bringing together our new ‘gentech’ models with decades of investment in imaging, typography, illustration and more to produce assets,”
Adobe says that artwork created using Firefly will have metadata indicating it is partially or wholly AI-generated. That’s a fair consideration and also a legal one that it hopes may address protests from artists on platforms such as ArtStation, voicing their discontent with the wave of AI-generated art.
Source: TechCrunch
As a fun fact, the first country to ban AI-generated media without watermarks is – China.
In the shadow of a potential ban in the in US and pushback in other countries, TikTok has refreshed its content moderation policies
The changed policies are not a major overhaul, most policies have stayed the same.
One newly expanded section, however, covers “synthetic and manipulated media ” — otherwise known as AI deepfakes.
Previously, TikTok only said that it would ban content that could “mislead users by distorting the truth of events or cause significant harm to the subject of the video.”
Now the company says all realistic AI generated and edited content must be “clearly disclosed” as such, either in the video caption or as an overlaid sticker.
It added that any synthetic media “that contains the likeness of any real private figure” or that shows a public figure endorsing a product or violating the app’s other policies, like hate speech, will be banned.
These tools have gained much popularity in recent months. They have been used to put public figures like President Joe Biden and former President Donald Trump in unexpected and unflattering scenarios.
The company is also publishing a list of eight “Community Principles” that it says will “shape our day-to-day work and guide how we approach difficult enforcement decisions.” Notably, the first two principles are “prevent harm” and “enable free expression.”
Source: The Verge
A new report from Gartner noted that many data and analytics functions “are receiving increased investment in areas such as data management, data governance and advanced analytics where more than 60 per cent of those surveyed reported increases. The average data and analytics budget is $5.41 million, and 44 per cent of data and analytics teams increased in size in the last year.
But are companies getting a real return on that investment?
The same report, notes that in a self-assessment survey, only 44 per cent of data and analytics leaders reported “their team is effective in providing value to their organization.”
The survey lists 17 different executive leadership traits which correlated high organizational and team performance.
Alan Duncan, Distinguished VP Analyst at Gartner said that “Top-performing CDAOs (Chief Data Analytics Officers) invest in their success by developing skills to thrive in ambiguous circumstances, articulate compelling value stories and identify products and services that can drive business impact.”
According to the report, CDAOs must prioritize strategy over tactics, as they serve multiple stakeholders across the business. In fact, 78 per cent of respondents rank corporate or organizational strategy and vision as one of the top three inputs to the D&A strategy.
The report also found that data and analytics leaders must be persistent to meet new demands, as they are tasked with a broad range of responsibilities.
It’s not an easy task, and the demands being placed on them reflect a growing confidence in CDAO’s abilities but this, in turn, means more work and pressure to deliver results.
All this, while 39 per cent of those responding also pointed out that the lack of available talent is also making the work more challenging.
Despite these demands, 43 per cent of top performing leaders reported that they were effective in committing time to their own professional development versus 19 per centof those who were rated as lower performers.
The survey was conducted online from September through November 2022 among 566 data and analytics leaders globally.
Source: Gartner
And a report from the Register today said that AWS was moving from diesel to hydrotreated vegetable oil or HVO, for its data centres in Europe. It reckons that this move may result in an 90 per cent reduction in greenhouse gas emissions when compared with diesel.
HVO is a biofuel made from vegetable oils and can even use waste cooking oil or residue oils according to AWS.
It has some great advantages over other substitute biodiesel fuels and it doesn’t require an modifications to the backup generators and remains stable even in very cold temperatures. And those generators are important – because if the power gets cut, you need a backup power supply to keep the servers and the cooling systems functioning.
Nobody wants to say that they fried their chips with the wrong kind of oil.
Source: The Register
That’s the top tech news for today. Hashtag Trending goes to air five days a week with the daily tech news and we have a special weekend edition where we do an in depth interview with an expert on some tech development that is making the news.
Follow us on Apple, Google, Spotify or wherever you get your podcasts. Links to all the stories we’ve covered can be found in the text edition of this podcast at itworldcanada.com/podcasts
We love your comments – good, bad or whatever. You can find me on LinkedIn, Twitter, or on Mastodon as @therealjimlove on our Mastodon site technews.social. Or just leave a comment under the text version at itworldcanada.com/podcasts
I’m your host, Jim Love, have a Wonderful Wednesday!
The post Hashtag Trending Mar.22-New GPT-4 spews more misinformation, TikTok clamps down on deepfakes and AWS moves from diesel to vegetable oil first appeared on IT World Canada.