Each week, we bring you 5 stories that resonated the most in our internal Slack channel #AI-news. We write the newsletter using various AI tools because we're an AI company and our marketing wants to move with the times too. ๐
Today you're reading the 54th issue in a row.
#1
Tasks: New ChatGPT Beta Feature for Planning and Reminders ๐ฏ.
OpenAI has launched a new beta feature called "Tasks" for ChatGPT ๐ฏ. This feature allows users to set reminders and plan one-time or recurring actions ๐. It is available to Plus, Team, and Pro subscribers, and users can enable it via the โ4.0 with scheduled tasksโ option in the model selection menu โ๏ธ.
Tasks let users schedule specific actions at designated times โฐ, such as daily weather updates, reminders about passport expiration, or simpler tasks like sharing jokes. Notifications about completed tasks are sent via the web, desktop, and mobile apps ๐ฑ, with a limit of up to 10 active tasks at a time ๐.
This step transitions ChatGPT from simple conversations to task management ๐งโ๐ป, an area typically associated with virtual assistants. It also aligns with OpenAIโs broader strategy to enhance features and justify subscription pricing. Although still in beta, OpenAI is gathering user feedback to improve reliability and address potential issues ๐ ๏ธ before the full rollout ๐. What do you think about it ๐ค?
ย
#2
DeepSeek V3 ๐ค: Chinese AI Model Surpasses Meta and OpenAI Giants ๐.
The Chinese start-up DeepSeek has unveiled its V3 model, which outperformed models from Meta and OpenAI in tests ๐.
With 671 billion parameters, the model was trained in just two months โฑ๏ธ, costing only $5.58 millionโsignificantly less than the budgets of major players ๐. This achievement highlights the progress of Chinese companies in AI development ๐ค, despite U.S. sanctions restricting access to cutting-edge chips ๐.
The release of pre-trained parameters (open weights) allows other developers to adapt the model for their own use ๐ ๏ธ, making it more accessible ๐. DeepSeek's success demonstrates that smaller companies can innovate rapidly, even when facing global technological restrictions ๐ช.
ย
#3
Apple AI Misstep โ: False Headlines Force the Giant to Disable Feature ๐ฐ๐ค.
Apple has temporarily disabled its AI news summarization feature ๐ค after it generated misleading and false headlines, including claims about the arrest of Israeli Prime Minister Netanyahu and fake BBC reports โ. These missteps sparked criticism from media outlets and press freedom organizations โ.
Part of Apple Intelligence ๐, the feature aimed to make news more accessible but instead raised concerns about accuracy and trustworthiness. Critics, including the BBC and Reporters Without Borders ๐ง, highlighted the risks of unreliable AI handling sensitive news content. Apple plans to fix the feature, add clear disclaimers that content is AI-generated, and relaunch it later ๐.
For now, the feature remains offline, underscoring the challenges of building trustworthy AI in the high-stakes realm of news ๐. Whatโs your take on this ๐ค?
ย
#4
Google Titans ๐ง : New AI for Processing Massive Datasets ๐๐,
Google has unveiled Titans, a new family of AI models designed to handle massive datasets ๐ (over 2 million tokens) with exceptional accuracy. Combining short-term memory ๐ง (for quick data processing) and long-term memory (to retain key patterns), Titans address GPT-4's challenge of processing lengthy contexts ๐.
These models excel in tasks such as text analysis, DNA sequencing ๐งฌ, and financial forecasting. Titans are top-tier for working with long documents or large datasets ๐๏ธ, making them ideal for pinpointing specific details (โfinding a needle in a haystackโ).
Industries like healthcare and finance stand to benefit significantly, from faster fraud detection โ ๏ธ to more accurate patient diagnostics and groundbreaking research. While Titans show great promise, they remain in development, with challenges like high computational costs and flexibility ๐ still needing resolution for broader adoption ๐.
ย
#5
Creators Profit from AI ๐ค: Google and OpenAI Pay for Unused Videos ๐ฅ๐ฐ.
Tech giants like Google and OpenAI are paying creators $1โ4 per minute for unused video footage to train their AI models legally and ethically ๐คโ๏ธ. High-quality content, such as 4K videos or 3D animations, holds the highest value. This move reflects efforts toward ethical data acquisition ๐ after criticism of unauthorized content scraping from the internet โ๐ธ๏ธ.
Mediators like Troveo AI simplify the process by managing and licensing content from creators ๐. This ensures companies get the data they need while creators receive fair compensation ๐ต. Agreements also protect creators' brands, preventing AI firms from replicating their styles or scenes ๐ฌ.
For creators, this is a unique opportunity ๐ to monetize unused content legally and transparently ๐ก. However, experts advise acting quickly โณ, as this opportunity may fade as the market evolves ๐.
Article on artificialintelligence-news.com
ย
