Dubai Police Introduces Self-Driving Security Patrol
In October 2023, at the Gitex Global 2023 event, Dubai Police showcased a new self-driving security patrol vehicle aimed at improving safety in residential neighborhoods. Lieutenant Rashed Binh Hadar from the General Department of Administrative Affairs introduced this eco-friendly electric patrol vehicle to the attendees.
The vehicle, with a battery life of up to 15 hours, cruises at speeds of 5 to 7 kmph, making it suitable for residential areas. The primary goal of this patrol is to augment security coverage in these zones. It is equipped with a high-tech camera capable of 360° imagery, along with smart technology and artificial intelligence to identify potential criminal activity, recognize individuals, and read car license plates.
Moreover, the patrol vehicle has direct communication technologies for real-time coordination with the Dubai Police's command and control center in the General Operations Department. Lieutenant Rashed also mentioned the inclusion of an onboard drone. The drone can venture into places the vehicle can't reach, maintaining wireless communication with the patrol and providing a more thorough security presence in residential areas.
GPT 4: Notable Speed Enhancements
Over the past few months, there has been a noticeable improvement in the speed of GPT 4, especially when compared to GPT 3.5. Data shows that the time it takes for GPT 4 to respond to requests is going down. Although the speed may vary depending on what you ask it to do, the overall trend shows that it's getting faster.
The article breaks down the time taken into three parts: round trip time, queuing time, and processing time, which change based on the complexity and length of your request. Interestingly, a longer request doesn't always mean a slower response. For example, a 2004 token request can get a response in 4.5 seconds if it's simple, while a 33 token complex request may take up to 32 seconds.
When comparing median request latencies (average times), they are quite consistent across GPT 3.5 and GPT 4, staying under 1 millisecond per token. However, when looking at the 99th percentile slower requests, the latencies have reduced by more than half in just 3 months, which is a significant improvement. While GPT 4 is still more expensive to use, it's no longer slower for the majority of requests.
YouTube Ventures into AI for Mimicking Musicians
YouTube is working on an AI tool that allows users to mimic the voices of famous musicians. The specifics are scant, but talks are ongoing with big music labels like Sony Music Entertainment to secure song rights for training the AI.
This follows a recent ban on a viral AI-generated song imitating Drake and The Weeknd on YouTube and other platforms. AI and music bring up copyright and licensing issues, particularly when replicating real individuals' voices. YouTube is addressing these hurdles by adopting new tech.
A few months back, it initiated AI music principles to ensure a balance between creativity and artist rights. This new feature raises questions about content boundaries, revenue sharing, and artists' rights if they disagree with how their voice is used. YouTube's journey in AI integration continues with recent features like automatic language dubbing, showing its commitment to blending AI with creativity while treading carefully on rights issues.
Mid Journey's High-Resolution AI Art
Mid Journey has introduced a new feature that significantly improves the resolution of its AI-generated art. Previously, images were capped at a resolution of 1,24 by 1,24 pixels, making them suitable for web use but not for larger prints.
Now, users can upscale images by 2X or 4X, achieving a maximum resolution of 4,096 by 496 pixels or 4K quality. This feature is versatile, as it can be applied to different aspect ratios. For a 16:9 image, 4X upscaling results in a resolution of 5,824 by 3,264 pixels. This means the images can now be printed at A3 size without losing clarity.
However, upscaling does have a downside. It takes longer, with 4X upscaling lasting 2 to 3 minutes. Additionally, since Mid Journey charges based on processing time, upscaling will consume a greater portion of the customer's monthly image generation allowance. They can opt to pay extra for more processing time.
Lastly, the company noted that in very rare cases, 4X upscaling might yield black or corrupted images, and blurry images won't be improved by upscaling.
AI Bots in the Spotlight on Shark Tank
On a recent episode of Shark Tank, Maurice Bachelor and Joel Griffith from Los Angeles presented their company, Bodit, aiming to alter negative perceptions about artificial intelligence bots. Their platform, including a website and mobile app, utilizes AI to streamline online chores like scheduling appointments or securing restaurant reservations.
An upgraded subscription offers an edge in quickly accessing concert tickets or sneaker releases, which stirred some debate due to the notorious reputation bots have in such scenarios. Michael Rubin, CEO of Fanatics and guest judge, acknowledged the problem bots create in buying up merchandise to likely resell at inflated prices.
Bachelor and Griffith argued that their technology actually assists in combating such cheating by providing individuals a chance to beat the bots. Mark Cuban was the first to bid, proposing $150,000 for a 20% stake. Michael Rubin countered, offering $500,000 for 15%, valuing the company higher. Eventually, a joint deal of $300,000 for 30% equity was struck, marking a significant milestone for Bodit in garnering the support of both billionaire investors to further their venture.
Global Expansion of Claude AI
Anthropic has expanded access to Claude AI to more individuals globally. Now, people in 95 countries can utilize Claude for assistance with their professional or regular chores. Since its initiation in July, Claude has attracted millions, owing to its vast memory, 100,000 token context window, and file upload capability, aiding in data analysis, enhancing writing skills, and even engaging with books and academic papers.
The expansion allows users in all these countries to use both the free version and Claude Pro, aiming to enhance their productivity. Anthropic is committed to increasing availability responsibly in the near future, with more updates to follow soon.
Microsoft CEO's Vision for AI: Accessibility and Innovation
In his annual letter, Microsoft CEO Satya Nadella shared the company's vision for a new era of AI, focusing on its accessibility, co-pilot integration, and dedication to innovation, financial growth, and responsibility across various sectors.
Amid global challenges, Nadella emphasized Microsoft's commitment to democratizing AI and driving innovation. He sees AI powered by natural language interfaces and a strong reasoning engine as a game-changer for software and businesses, aiding productivity and addressing global issues.
Nadella introduced Co-Pilot, an AI companion in Microsoft's products designed to assist users in different tasks. He highlighted three core commitments: maintaining innovation, achieving robust financial performance with a revenue of $211 billion in 2023, and ensuring responsible AI development.
NYC Mayor's Multilingual Robocalls Initiative
NYC Mayor Eric Adams employed AI technology for robocalls to converse with residents in various languages. Although he only speaks English, he sent out numerous calls in Spanish, Yiddish, Mandarin, Cantonese, and Haitian Creole to potentially engage with more voters.
This initiative, however, stirred concerns about its ethical stance, as it gives a false impression of the mayor's linguistic capabilities. Critics argue that it's misleading and an inappropriate use of taxpayers' money.
During a recent press briefing on the city's artificial intelligence action plan, Adams and the city's Chief Technology Officer discussed the broader applications of AI, like the new tool My City Chatbot to improve city services. Despite the criticism, Adams justified the multilingual robocalls as a means to connect with non-English-speaking residents, aiding in the promotion of local events and resources. The practice, especially making it sound like Adams speaks these languages, has been labeled as unethical and a part of a creepy vanity project by some observers.
Nvidia's New AI Agent: Eureka
Nvidia has introduced a new AI agent named Eureka, which is designed to train robots automatically by generating algorithms. This development was shared in a research paper where it was revealed that Eureka utilizes GPT 4's natural language processing capabilities along with reinforcement learning to assist robots in mastering complex tasks.
Remarkably, Eureka was able to train a robot hand to spin a pen like a human, open drawers, catch balls, and perform 30 other varied tasks without any human intervention. Ana Anandkumar, Nvidia's Senior Director of AI Research, mentioned that Eureka is a progressive step in crafting new algorithms that merge generative and reinforcement learning to tackle challenging tasks.
Eureka operates in a robotic simulation environment, creating reward functions or algorithms that guide robots to learn through a repetitive process of trial and error. To test this, Nvidia employed its Isaac Gym, a physics simulation platform. Furthermore, Nvidia has shared Eureka's code and benchmarks openly for developers to build upon this research and conduct their own experiments using the simulator platform.
That concludes this week's AI news. Stay tuned for more updates on the latest advancements in artificial intelligence.
0 Comments