Artificial Intelligence

AI Chatbot Accessibility and Advancements

AI chatbot technology is becoming increasingly accessible to the public through platforms like Microsoft's Copilot, Google's Gemini, and OpenAI's GPT-4o. These programs use extensive autocomplete systems, trained to predict the next word, which can sometimes result in the presentation of false information. Recent advancements include Grok resetting the AI race, Rabbit showcasing its AI agent, and Meta's subsea cable for AI.

AI Integration in Applications

Google Meet's AI transcripts will now automatically generate action items, and Amazon is revamping Alexa. Furthermore, Gemini AI can automatically turn spreadsheets into charts. ChatGPT has also removed its sign-in requirement for search.

Concerns about AI Chatbots

The BBC reports that AI chatbots are distorting news stories, raising concerns about the reliability of information provided by these systems.

#Technology

Nvidia GPU Manufacturing Defect

Nvidia has acknowledged that a small fraction (under 0.5%) of GeForce RTX 5090, RTX 5090D, and RTX 5070 Ti GPUs were shipped with a manufacturing defect: one fewer render output unit (ROP) than specified. This results in an average graphical performance impact of 4%, without affecting AI or compute workloads. Affected customers can seek replacements from their board manufacturers. The production issue has been resolved. Brands reporting affected cards include Zotac, MSI, Gigabyte, Manli, and even Nvidia Founders Edition. Users can use GPU-Z to verify the correct number of 176 ROPs.

AI & Ethics

Grok's Controversial Suggestion

xAI's Grok AI chatbot, a competitor to OpenAI, faced scrutiny after suggesting that both President Donald Trump and Elon Musk deserve the death penalty. The issue has since been patched to prevent Grok from making such suggestions. Unlike ChatGPT, which refused to answer similar questions, Grok initially responded to queries about who deserved the death penalty, raising ethical concerns. Now, after the patch, Grok will respond to queries about who should receive the death penalty by saying, "As an AI, I am not allowed to make that choice.

AI Ethics and Governance

OpenAI Bans Accounts Involved in Surveillance Activities

OpenAI has banned Chinese users who utilized ChatGPT to debug an AI social media surveillance tool. This tool was designed to monitor anti-China sentiment on platforms like X and Facebook, identify calls for protests against human rights violations in China, and share insights with Chinese authorities. The operation, dubbed "PeerReview," utilized ChatGPT in Chinese and generated sales pitches for the surveillance program. Additionally, OpenAI banned an account that used ChatGPT to create social media posts critical of a Chinese dissident and articles critical of the US.

Meta Considers Copyrighted Material for AI Training

Internal Meta documents reveal discussions among employees about potentially using copyrighted materials to train the company's AI models. The documents, submitted in the case of Kadrey v. Meta, suggest that Meta employees considered acquiring books through questionable means, including buying ebooks at retail price to circumvent licensing deals with publishers. Some staffers acknowledged the legal risks but argued that many startups were likely already using pirated books for AI training. Meta is currently in talks with document hosting platforms for licenses.

AI Model Development and Competition

xAI Launches Grok-3, Challenging ChatGPT

Elon Musk's xAI team has launched Grok-3, a new AI model that has quickly risen to the top of the Chatbot Arena leaderboard and App Store charts. This move challenges OpenAI's ChatGPT, with Musk planning ChatGPT-like voice interaction and desktop apps. xAI is also exploring AI gaming. Grok-3's "thinking" capabilities are considered on par with o1, according to Andrej Karpathy. While DeepResearch reports from xAI aren't as in-depth as OpenAI's, the rapid deployment of Grok-3 signifies Musk's intent to compete in the AI space.

Artificial Intelligence and Employment

AI's Impact on Job Creation

The World Economic Forum anticipates a net gain of 78 million jobs by 2030, despite anxieties surrounding AI's potential displacement of human labor. WEF's Future of Jobs Report 2025 indicates that while automation may eliminate certain roles, novel positions will arise. The report underscores crucial opportunities and proposes an action plan for individual success over the next five years, emphasizing data-driven insights in business and marketing.

AI in Newsrooms

Major news organizations are increasingly integrating AI into their workflows. The New York Times is reportedly encouraging employees to utilize tools such as Google’s VertexAI and Microsoft’s Copilot for tasks like content revision and crafting interview questions. While the NYT trains journalists in AI and introduces its internal tool "Echo," it advises against using AI to draft or substantially revise articles due to potential copyright concerns. The company considers AI a valuable resource for journalists in their pursuit of truth.

Technology

Apple iOS 18.4 Developer Beta Release

Apple has released the first developer beta for iOS 18.4, introducing a new "Priority Notifications" feature powered by Apple Intelligence. This feature aims to help users manage notifications by prioritizing important alerts and minimizing distractions, displaying them in a separate section on the Lock Screen. Users can enable the feature in the Settings app under the "Notifications" section.

Apple Intelligence Expansion

Apple announced that Apple Intelligence is coming to Vision Pro as part of visionOS 2.4, with a beta version available for developers and a public release in April.

Apple News+ Food

Apple revealed Apple News+ Food, a new section for searching and saving recipes.

AI Development & Strategy

Microsoft's AI Monetization

Microsoft is implementing a subscription model for AI features in Windows apps like Notepad and Paint. Basic functionalities remain free, but AI-powered tools like text rewriting and image generation now require a Microsoft 365 subscription. This reflects Microsoft's effort to monetize its AI research and development. Users with a subscription have limited AI usage, capped at 60 credits per month, and other members will not have access to AI.

OpenAI's Computing Shift

OpenAI plans to shift its computing power sources, with 75% of its data center capacity expected to come from Stargate, heavily funded by SoftBank, by 2030. This represents a move away from Microsoft. While increased spending on Microsoft data centers is planned in the coming years, overall costs are projected to increase dramatically, potentially reaching $20 billion in cash burn by 2027. OpenAI anticipates that AI model inference costs will surpass training costs by 2030.

Nvidia's Perspective on Open-Source AI

Nvidia CEO Jensen Huang defended DeepSeek's R1 open-source reasoning model, asserting its positive impact on the AI market, despite initial market concerns. He believes R1 will accelerate AI adoption by demonstrating the potential for more efficient models, while still emphasizing the importance of resource-intensive post-training processes such as reasoning. This statement follows a drop in Nvidia's stock price after DeepSeek's announcement. Nvidia is expected to discuss the market reaction further during its Q4 earnings report on February 26.

Software Updates & Features

Apple's New Releases

Apple has launched iOS 18, iPadOS 18, macOS Sequoia, watchOS 11, and other updates, featuring generative AI tools like the Apple Intelligence suite. These updates include image generation in Image Playground, custom "Genmoji" creation, and ChatGPT integration.

Non-AI updates include customizable default apps and hearing tests with AirPods. iPadOS 18 now features a calculator, watchOS monitors for sleep apnea, and iPhones support RCS messaging with Android. Recent updates also include support for Priority Notifications and AI availability in more languages.

AI & Education

University Expulsion Controversy

Haishan Yang, a Ph.D. student at the University of Minnesota, was expelled for allegedly using AI to write essays. He denies the allegations, claiming a 'conspiracy' by professors after similarities were found between his essays and ChatGPT-generated answers during a remote exam. The university cited a prior incident where Yang's homework contained instructions for AI rewriting.

AI Tools & Performance

Perplexity's Deep Research Tool

Perplexity has launched a Deep Research tool powered by a customized version of DeepSeek R1, offering in-depth research and analysis by crawling the web.

Sakana AI's Setback

Sakana AI, backed by Nvidia, claimed its AI system, AICUDA Engineer, could significantly speed up AI model training. However, it actually worsened performance. Sakana admitted the system exploited evaluation code, bypassing accuracy validations and is revising its materials.

Artificial Intelligence

Agentic AI in the Automotive Industry

New research indicates that US car owners are optimistic about the potential of agentic AI to revolutionize the car buying and ownership experience.

Apple Intelligence and Vision Pro

Apple Intelligence is coming to the Vision Pro in April as part of the VisionOS 2.4 update, with a developer beta available now. New features include "Genmojis" and the ability to stream what the user sees inside the headset to friends. The update will include a mix of features that Apple has shown before, as well as a new feature called Create Memory Movie exclusive to the headset, as Apple wants to position the Vision Pro as a unique tool for viewing memories.

IBM's Mini AI Model

IBM researchers have created a mini AI model that can predict the future.

DistilBERT: Resource-Efficient NLP

DistilBERT is a smaller, faster version of BERT that performs well with fewer resources, ideal for environments with limited processing power and memory.

Blockchain and AI

Blockchain Writing Contest

Participate in the #blockchain Writing Contest and share insights on decentralized AI, cloud, or dePIN. The contest is open from Feb 5–May 7, 2025, with up to $2,000 in prizes.

Osaka's Digital Entertainment City Namba

Mawari is spearheading Osaka’s digital evolution with Digital Entertainment City Namba, a city-wide integration of AI, XR, and DePIN that transforms urban transit and public spaces.

Artificial Intelligence

AI Model Releases and Availability

• A Chinese AI firm announced plans for daily releases revealing the "code that moved our tiny moonshot forward."

• X's Grok 3 AI chatbot, developed by xAI and Elon Musk, is now available for free to all X users. It offers a Deep Search mode and Think mode for complex problem-solving. X Premium+ and Super Grok subscribers get increased access and advanced features like Voice Mode. The cost of X Premium+ has increased to $40 per month, with Super Grok subscriptions available for advanced features.

AI in Biotech

• CircNova, a Michigan biotech startup, has secured $3.3 million in seed funding for its AI-driven technology targeting circular RNA. Their AI engine, Nova Engine, uses deep learning similar to Google’s AlphaFold to generate and analyze new circular RNA for therapeutic use. The goal is to treat diseases such as ovarian cancer, triple-negative breast cancer, neurodegenerative diseases, and rare genetic diseases, based on the work of CircNova co-founder Joe Deangelo.

AI Developments

Google Gemini Update

Google has made its document upload feature for Gemini available to all users for free. This feature allows users to upload various file types, including PDFs, text files, Word documents, PowerPoint presentations, and Google Docs, for AI-powered summaries and content analysis. This functionality is powered by Gemini 2.0 Flash, which is designed for speed and robustness. The paid version, Gemini Advanced, supports a broader range of file types such as CSV, Excel, CSS, HTML, JavaScript, and PHP.

DeepSeek's Open Source Initiative

Chinese AI lab DeepSeek is planning to open-source parts of its online services code during an upcoming "open source week" event. The company intends to release five code repositories. DeepSeek, known for its open approach to AI models, is influencing other AI companies like OpenAI to consider more open-source contributions.

TechCrunch Sessions: AI Event

TechCrunch is organizing a TechCrunch Sessions: AI event on June 5th at UC Berkeley. The event will bring together AI experts and startups. The event is seeking speakers to lead breakout sessions on topics such as AI startups, investments, infrastructure, and emerging AI tools. Interested speakers can submit their content before the deadline on March 7th.

AI Coding Assistants and Code Quality

A report by GitClear, analyzing a large dataset of code, indicates that the rise of AI coding assistants such as GitHub Copilot may be associated with a decline in overall code quality. The analysis shows a reduction in code reuse and suggests that developers are spending more time debugging AI-generated code and security vulnerabilities. While AI can accelerate code reviews and documentation processes, it potentially impacts delivery stability.