| Here's a look at today's AI briefing: - Microsoft blocks certain Copilot AI prompts to prevent inappropriate imagery.
- Florida students charged with felonies over AI deepfakes.
- Survey: Employers would raise salaries by up to 44% for AI-skilled workers.
Thank you! Beth
p/beth-duckett | |
| 1 | | Microsoft has blocked some prompts in its Copilot AI tool to prevent it from generating certain inappropriate images. The move comes after a Microsoft AI engineer warned the FTC about the tool, citing its potential to create unsafe, violent, and sexual imagery. More: - Copilot's image-generation AI now blocks prompts such as "pro choice," "pro life," and "four twenty," according to CNBC.
- It also no longer generates images of children playing with assault rifles, which it could do previously.
- The system flags the prompts as potential policy violations, warning that repeated attempts could result in an automatic suspension.
- CNBC found Copilot could still create violent images with prompts like "car accident" and generate copyrighted content such as Disney characters.
- Microsoft said it's continuously tracking and updating Copilot to improve safety filters and present misuse.
Zoom out: - Shane Jones, a Microsoft principal software engineering manager, sent letters to the FTC and Microsoft's board this week warning about Copilot's flaws and security risks.
- He has urged Microsoft to withdraw Copilot's image creator from the public until stronger safeguards are in place.
|     | |
| 2 | | Two Florida middle-schoolers were charged with felonies for allegedly creating AI deepfake nudes of their classmates, according to Wired. Reports noted it may be the first case involving criminal charges levied against students for AI-generated nudes. More: - The students, aged 13 and 14, are accused of using an unidentified AI app to create nonconsensually generated images of their peers aged 12 and 13.
- The students were arrested in December and charged with third-degree felonies for distributing deepfake sexually explicit images without consent, which is banned under Florida law.
- While most states have laws against revenge porn, only a few have passed legislation specifically targeting AI-generated explicit content.
Zoom out: - Meanwhile, the Beverly Hills Police Department in California is investigating a case involving middle school students sharing AI-generated nude images with other students' faces.
- The school district's board voted this week to expel five students implicated in the scandal.
| | | |
| 3 | | Employers are willing to raise salaries by up to 44% for AI-skilled workers, especially in the IT and R&D sectors, according to a survey commissioned by Amazon Web Services (AWS). 
What happened: The study, carried out by Access Partnership for AWS, surveyed 4,664 employers and 14,896 workers in nine Asia-Pacific countries, including Australia, India, and Japan. What the numbers show: The survey showed employers are willing to hike salaries by up to 44% for AI experts in IT and 41% in R&D. They would pay 39% extra for AI talent in sales, marketing, and business operations, and 37% more for finance workers with AI skills. On average, workers with AI skills could expect a 33% raise. Employers in India were willing to offer the highest raise at 54%, the top among surveyed countries. Relevance: AI is already transforming industries, raising the demand for AI-related skills and talent as companies seek a competitive edge. Facing a shortage of AI experts, companies are ready to pay top dollar for skills in areas like GenAI, prompt engineering, Python, natural language processing, and more. | | | |
| 4 | | AI fraud detection startup Inscribe has laid off nearly 40% of its workforce, affecting dozens of employees. Most of the affected roles were in go-to-market and operations, CEO co-founder Ronan Burke told TechCrunch. More: - Inscribe's board reportedly recommended the layoffs after the company missed revenue goals for more than a year.
- Burke noted that in 2023, Inscribe's fintech clients faced higher interest rates and uncertainty for consumers and businesses.
- Meanwhile, last year's AI advancements have bolstered opportunities for the financial sector, he said.
- In Q4, Inscribe adopted a new product strategy to match these industry shifts, with a major launch planned for this year, Burke said.
Zoom out: | | | |
| |
|
| |
| 6 | | Weekly roundup — The top AI stories you might have missed this week: - OpenAI responded to Elon Musk's lawsuit accusing the company of prioritizing profit over its mission to develop open AI that benefits humanity. OpenAI shared emails written by Musk, an early OpenAI co-founder, where Musk supported raising billions in funding and merging OpenAI with Tesla.
- A federal jury charged an ex-Google engineer with stealing AI trade secrets for two Chinese companies he secretly worked for. Linwei Ding is accused of stealing more than 500 secret files on Google's AI, mainly concerning its TPU chips used in AI operations.
- U.S. chipmaker AMD will need an export license to sell its AI chip specifically made for Chinese customers. While AMD sought approval from the U.S. Commerce Department for the chip, regulators decided it would still need an export license.
- Anthropic announced its most advanced family of AI models, Claude 3, rivaling Google and OpenAI's models. Anthropic's latest models have both text generation and image analysis capabilities. The largest, Opus, outperformed GPT-4 and Google's Gemini Ultra in benchmark tests involving math, knowledge, and reasoning.
- A BBC Panorama investigation uncovered many AI-generated photos of Donald Trump appearing alongside Black individuals. The fake images appear to have been created by some Trump supporters and have no ties to the former president's campaign.
- Popular AI-powered image creators can be used to create misleading photos about elections and voting, even with built-in safeguards, according to a study by the Center for Countering Digital Hate (CCDH). In the study, the nonprofit was able to generate fake and "convincing" images of Joe Biden in a hospital bed, Donald Trump in a jail cell, and ballots in a dumpster.
| | | |
| 7 | | QUICK HITS - Sales teams at MongoDB, Warmly, and User Evidence are seeing 56% more pipeline from this simple framework.*
- Google says its latest Pixel 8 smartphone cannot run its smallest Gemini AI model due to "hardware limitations." However, the on-device AI model does power various features in the more expensive Pixel 8 Pro phone.
- Broadcom said it expects $10B in AI-related chip revenue this year. The U.S. chipmaker creates networking chips for managing AI data and helps clients design their own AI chips.
- LinkedIn's premium subscriptions surged by 25% in 2023, generating $1.7B. The paid service includes AI features for better job search and recruitment.
- Microsoft is introducing "Speak For Me," a tool to help people with ALS and speech disabilities communicate with custom neural voices, later this year.
- U.S. presidential candidate Jason Palmer beat President Biden in American Samoa's caucus. Palmer attributed the win in part to his campaign's use of generative AI to communicate with voters and answer questions via a PalmerAI chatbot.
*This is a sponsored post. | | | |
| Upcoming Events | | MAR
27 | | Discover how generative AI can speed up your enterprise, expand your business model, and improve your value proposition. Network with hundreds of AI leaders.* | | | | | | * This is a sponsored event | | | |
 | AI and technology writer | | Beth is a contributing editor and writer of Inside's AI and Tech newsletters. She has written for publications including USA Today, the Arizona Business Gazette, and The Arizona Republic, where she received recognition with a Pulitzer Prize nomination and a First Amendment Award for collaborative reporting on state pension cost increases. You can reach her at Beth.Duckett@yahoo.com. | | This newsletter was edited by Aaron Crutchfield | |
| |
| |
Comments
Post a Comment