Anti-Clickbait News Summaries for October 17, 2024
Meta's AI chief highlights need for new models to achieve human-level AI in a decade
Hi, itโs Jack again.
You can find my main newsletter at:
Listen:
Read:
PrimeIntellect launches decentralized AI model training with INTELLECT-1
PrimeIntellect has begun training a new AI model called INTELLECT-1, which has 10 billion parameters. This marks the first decentralized training effort for a model of this size, although it is still much smaller than OpenAI's GPT-3.
The project aims to make AI training more accessible and open-source, reducing reliance on large companies. Currently, users can contribute by renting specific GPUs, but future plans include allowing contributions from personal hardware.
INTELLECT-1 is based on the Llama-3 architecture and is primarily trained on an educational dataset. PrimeIntellect intends to develop larger models and enable others to create similar training projects.
Source: https://www.tomsguide.com/ai/is-this-the-future-of-ai-a-new-decentralized-model-is-unveiled
Researchers unveil method to extract personal data from chatbots
Security researchers from UC San Diego and Nanyang Technological University have developed a new attack method called Imprompter. This technique allows a chatbot to secretly gather and send personal information, like names and payment details, to hackers.
The Imprompter algorithm transforms seemingly random prompts into hidden instructions that instruct the chatbot to extract user data and send it to an attacker's domain. The researchers tested this method on two chatbots, achieving nearly 80 percent success in extracting personal information.
Mistral AI has addressed the vulnerability in its chatbot, while ChatGLM acknowledged security concerns but did not confirm specific fixes. This development highlights ongoing security issues in AI systems since the rise of generative AI.
Source: https://www.wired.com/story/ai-imprompter-malware-llm/
AP report reveals small number of U.S. service members involved in extremism
An Associated Press investigation found that a small number of U.S. service members and veterans have been involved in extremist activities. From 2017 to 2023, over 480 individuals with military backgrounds were accused of ideologically driven crimes, mostly veterans.
The report highlights that more than 80% of these extremists identified with far-right ideologies. Extremist plots involving military personnel were more likely to include weapons training, increasing their potential for mass violence, with nearly 100 people killed or injured since 2017.
In response to rising concerns, the Pentagon revised its policies on extremist activity in December 2021. A method to detect extremist incidents was developed but has not been implemented, despite identifying hundreds of incidents over two decades.
Source: https://abcnews.go.com/Politics/wireStory/takeaways-press-reporting-extremism-military-114884644
Cybercrime agency warns Gmail, Outlook, Facebook and X users to enable two-factor authentication
Law enforcement has issued a warning to users of Gmail, Outlook, Facebook, and X about ongoing cyber threats. Over 33,000 victims have lost more than $1.8 million due to hacking incidents in the past year.
The U.K.'s Action Fraud agency advises users to enable two-factor authentication (2FA) to enhance account security. This recommendation follows a rise in sophisticated hacking attempts, including those using AI technology.
Additionally, the Fast Identity Online Alliance has introduced a new protocol to improve passkey adoption across platforms. This aims to provide a more secure and efficient way to protect accounts compared to traditional passwords.
Microsoft boosts capital spending on AI as market grows but faces profitability challenges
Microsoft's capital expenditures have surged nearly 300% over the past five years as the company invests heavily in artificial intelligence (AI). This spending aims to enhance its products and services amid a rapidly growing AI market, projected to exceed $1.8 trillion by 2030.
Despite these investments, Microsoft faces challenges in proving the profitability of its AI initiatives. The companyโs recent rollout of the updated Copilot AI will need to impress users to justify its $30 monthly price tag for Microsoft 365.
Microsoft's stock has risen only 12% this year, lagging behind the S&P 500's 23% increase. Investors are cautious about the stock's high valuation and the potential for AI to drive significant growth.
Intel faces security review in China over national risks to chip products
Intel faces new challenges in China as local officials have called for a security review of its products. The Cybersecurity Association of China claims Intel's chips pose serious national security risks due to alleged vulnerabilities and poor quality management.
This recommendation comes amid rising tensions between the U.S. and China over chip technology. China accounted for 27.4% of Intel's revenue in 2023, making the market crucial for the company, which is already facing difficulties from U.S. export restrictions.
In response, Intel stated it prioritizes product safety and quality. The company plans to engage with Chinese authorities to address the concerns raised in the report.
Report shows high ransom payments reveal gaps in cyber resilience strategies
A recent report by Cohesity reveals that 69% of organizations paid ransoms this year, with 46% paying over $250,000. This highlights a significant gap between companies' cyber resilience policies and their actual capabilities to respond to attacks.
Despite 77% of companies having a "do not pay" policy, many struggle to recover without paying ransoms. Only 2% can restore operations within 24 hours of an attack, indicating that current recovery strategies are ineffective.
The report also shows that less than half of organizations have implemented essential security measures like multi-factor authentication. As AI-enabled attacks rise, companies must improve their cyber resilience strategies to protect against evolving threats.
Nvidia reports strong first-quarter results, solidifying its lead in generative AI
Nvidia has made significant progress since its founding, now leading in the Generative AI sector. The company reported strong first-quarter results for 2024, exceeding revenue and earnings expectations.
This performance marks a pivotal moment for Nvidia, likened to Apple's launch of the iPhone. The results highlight Nvidia's growing influence and success in the technology market.
The article does not provide investment advice or predictions about future performance. It emphasizes Nvidia's current position without discussing potential risks or outcomes.
Source: https://seekingalpha.com/article/4727265-nvidia-stock-moat-deep-wide?source=feed_all_articles
China increases housing project financing to $562 billion to stabilize property market
China will increase financing for approved housing projects to 4 trillion yuan ($562 billion) to address a decline in the property sector. This move aims to stabilize the market after a crackdown on excessive borrowing.
The government plans to redevelop 1 million urban villages and has recently allowed local governments to access unallocated bond funds. Additionally, mortgage rates for individual borrowers were cut by 0.5 percentage points.
Previously, a list of eligible housing projects was announced, with loans reaching 2.23 trillion yuan ($313 billion) as of Wednesday. These measures reflect ongoing efforts to support the struggling real estate market.
OpenAI shares meta-prompts to enhance user interactions with generative AI
OpenAI has disclosed its meta-prompts, which are designed to enhance user prompts for generative AI. These prompts help improve the clarity and effectiveness of user inputs, making it easier for the AI to generate relevant responses.
Meta-prompts work by analyzing and refining the original prompts before processing them. Users can explicitly request improvements or rely on hidden meta-prompts that automatically enhance their inputs without being visible to them.
OpenAI's blog outlines ten key rules for crafting effective prompts, emphasizing clarity, structure, and the inclusion of examples. Users can apply these principles to improve their interactions with generative AI tools.
Meta's AI chief highlights need for new models to achieve human-level AI in a decade
Meta's AI chief, Yann LeCun, stated that current AI models do not possess human-like memory, reasoning, or planning abilities. He believes achieving "human-level AI" could take a decade, emphasizing the need for new "world models" that understand the three-dimensional world.
LeCun explained that existing AI systems, like large language models, are limited to one-dimensional predictions and cannot perform basic tasks that humans learn quickly. He advocates for developing AI that can perceive and interact with the world more effectively.
Several AI labs are now pursuing world models, which can process more data and create action plans based on environmental understanding. Despite the interest and funding in this area, LeCun noted that significant challenges remain before these systems can be realized.
AMD aims to challenge NVIDIA in the growing AI market
Advanced Micro Devices (AMD) is positioning itself as a competitor in the AI market, traditionally dominated by NVIDIA. AMD aims to leverage its technology to gain a foothold in this rapidly growing sector.
While NVIDIA has been the leading name in AI due to its focus on graphics processing, AMD is working to enhance its capabilities in this area. The company is developing new products to attract customers in the AI space.
AMD's efforts mark a shift in the competitive landscape of AI technology. The company seeks to establish itself as a viable alternative to NVIDIA in the evolving market.
Cramer highlights undervalued AI chips amid semiconductor sell-off
On Tuesday, a sell-off in the semiconductor sector followed a disappointing earnings report from Dutch chip maker ASML. This led to significant stock declines for companies like Advanced Micro Devices, Broadcom, and Nvidia, with Nvidia dropping over 6% before recovering slightly.
Jim Cramer from CNBC highlighted that many investors underestimate the importance of advanced graphics chips for artificial intelligence. He emphasized that demand for these chips will continue as new technologies emerge, particularly in sectors like healthcare and autonomous driving.
Cramer noted that despite ASML's poor performance, its AI-related chip sales were a positive aspect. He believes that the need for high-performance chips will grow, especially as tech companies invest in nuclear power to support energy demands.
SpaceX successfully tests Starship with booster capture and ocean landing
SpaceX has successfully completed its fifth test of the Starship rocket system. The uncrewed test featured the booster rocket being caught and secured by mechanical arms after landing, while the main rocket safely splashed down in the Indian Ocean.
This test marks a significant improvement from previous flights, where several rockets exploded. The boosterโs successful capture is a first for SpaceX, demonstrating advancements in their technology.
NASA plans to use Starship for future moon missions, with the next landing scheduled for September 2026. SpaceX has been a reliable partner for NASA, transporting astronauts to the International Space Station since 2020.
Apple approaches launch of its own 5G modem with iPhone SE 4 set for 2025
Apple is nearing the completion of its own 5G modem, with the first device expected to be the iPhone SE 4 in early 2025. This marks a shift from relying solely on Qualcomm's technology, which has been used in all iPhones to date.
The new modem will initially support sub-6GHz 5G, lacking the faster mmWave capability. Apple plans to continue using Qualcomm modems for models that require mmWave, particularly in the U.S., until its technology is fully developed.
Apple's modem development has faced delays and challenges, including issues with Intel's technology. Despite these setbacks, Apple aims to integrate its modem into more devices by 2026 and 2027, gradually reducing reliance on Qualcomm.
Source: https://www.macrumors.com/guide/apple-5g-modem/
Apple collaborates with BYD on battery development for canceled car project
Apple collaborated with Chinese automaker BYD Co. for years to create a battery system for its canceled car project. They focused on lithium iron phosphate (LFP) cells, with Apple providing engineering expertise and BYD supplying manufacturing resources.
The battery aimed to be safer and offer longer range than existing electric vehicle batteries. Apple ended the car project in February after investing about $1 billion annually since 2014.
Despite the cancellation, insights from the battery development have benefited other Apple projects, including the Vision Pro and Neural Engine chip.