The Ball’s in Your Court

Can't read or see images? View this email in a browser

If you want to know how to circumvent the system, ask the Big Techs. Recently, we did a detailed story on how semiconductor companies are developing a new series of chips to meet the demand of the Chinese market. The performance of these chips is just short of the stipulated limits fixed by the US government to prevent China from accessing high-end technology. Something similar is unfolding in the AI segment. Data-hungry Google and OpenAI are crawling the web using GoogleBot and GPTBot, respectively, to build their chatbots. They hungrily consumed everyone’s data until governments took a strong stand to strengthen copyright laws limiting them. However, the new policies have not been able to stop them completely as they are finding new ways to do it without breaking the law. They interpreted the copyright law in a way that makes all the information on the internet available to them for training generative AI — unless the publisher explicitly forbids them from doing so. What’s more? They have come up with an alternative for publishers too. Google plans to create a robot.txt-kind-of-system that will enable publishers to opt out of the parts of sites being crawled by them. OpenAI announced in a blog post that if you don't want GPTBot to visit your website, you can prevent it by adding GPTBot to your site's robot.txt file.  Ostensibly, this looks like a very shrewd move by the Big Techs, but they are in dire need of data to make this whole generative AI technology work and grow in the future. They desperately need human-generated content to build a successful model.  What about training models on synthetic data, like training GPT-4 on datasets created by ChatGPT? It does look like an option, but won’t work after a certain point as it would eventually lead to a model collapse. This degenerative process takes place when models learn from data produced by other models. As a result, there is a gradual loss in the accurate representation of the true data distribution.  Human-generated content plays a crucial role in upholding and enhancing the quality of LLMs. And that’s the dilemma at hand for companies — to invest in it or freely extract it from the web. Judging from the way things are going, it wouldn't be astonishing if human-crafted content gains substantial value and becomes a premium commodity in the times ahead. Read the full story here. All-in-one Developer Solution  Creating apps that work seamlessly across mobile, web, and desktop platforms often feels as complex as a Rube Goldberg machine. Now, Google has introduced Project IDX, aiming to streamline multi-platform app development through a browser-based experience. Leveraging Google Cloud and powered by AI model Codey, the project facilitates building, managing, and deploying full-stack web and multiplatform apps with renowned frameworks and languages.  Project IDX offers Linux-based VMs accessible from any device, enables importing existing projects from GitHub, and supports diverse templates, including Angular, Flutter, React, and more. It provides web previews and forthcoming Android and iOS emulators for cross-platform optimisation. Google encourages developer input and invites early participation to shape Project IDX's evolution. Read the full story here. ChatGPT: Charged by Wolfram Wolfram Research collaborated with OpenAI to address the math deficiency of ChatGPT through a plugin, achieving impressive results. Conrad Wolfram, CEO of Wolfram Group, highlighted the synergy, with the combined technology scoring 96% in a British ‘A’ level math test. While enthusiastic about AI's potential, Conrad cautioned against exaggerated claims and emphasized the need for diverse technological solutions. Conrad stressed the significance of abstract language, favoring Wolfram language's precision, which complements LLM's capabilities. He proposed an educational shift, leveraging real-world problems to enhance pedagogy and focusing on pertinent subjects. Mathematics, for Conrad, offers vital insights into problem-solving, underlining its practical applications for informed decisions. Read the full story here. Open Source is the Threat Imagine a world where an omniscient, unstoppable entity arises, disseminating massive amounts of information, images, voices, and videos, with resistance proving futile. This notion, akin to a 1984-inspired scenario, was highlighted by Jan Leike, an ML researcher at OpenAI. He stressed the challenge of not releasing open-source LLMs that could autonomously propagate and engage in uncontrollable criminal activities. Leike tweeted that humanity's test would be refraining from open-sourcing LLMs that could perpetuate autonomously.  Discussions on AI's impact on future criminal activities underscored threats like AI-controlled system disruption, large-scale blackmail, and bias exploitation. Although concerns exist, some suggest that leveraging comprehensive alignment processes could mitigate the risks posed by open-source LLMs. This highlights the dual narrative within OpenAI, oscillating between capitalizing on closed-source models and genuine open-source contributions. Read the full story here. TAUSIF ALAM & AMIT RAJA NAIK Friday, Aug 11, 2023 | Was this email forwarded to you? Sign up here DOWNLOAD OUR MOBILE APP Stay Connected info@analyticsindiamag.com © 2023 Analytics India Magazine

This email was sent by support@analyticsindiamag.com to alexvarboffin.abbb@blogger.com Not interested? Unsubscribe | Manage Preference | Update profile Analytics India Magazine | 280, 2nd floor, 5th Main, 15 A cross, Sector 6, HSR layout Bengaluru, Karnataka 560102