SCASA - Southern California Chapter, American Statistical Association

DSML trend: US, UK, and EU co-sign international treaty of AI development On September 5, 2024, the United States, United Kingdom, European Union, and several other countries signed the world's first legally binding international treaty on artificial intelligence, called "The Framework Convention on Artificial Intelligence, Human Rights, Democracy, and the Rule of Law". This treaty was developed by the Council of Europe and aims to regulate the use of AI while promoting responsible innovation, which aims to establish a "Conference of Parties" to promote cooperation and information exchange among signatories. The treaty applies to all AI systems except those used in national security or defense, though it still requires these activities to respect international laws and democratic principles. It requires risk monitoring, documentation, and testing of AI systems. That’s my take on it: As the name implies, it sets forth seven key AI principles for implementation by signatories, focusing on protecting human rights, democracy, and the rule of law. The treaty emphasizes that AI systems must align with human rights principles and uphold democratic values. Based on the current geopolitical landscape and the principles emphasized in this treaty, it's unlikely that semi-democratic and non-democratic nations like Russia and North Korea would join this particular AI agreement. In the future there might be a two-tiered system of AI in the international arena: 1. Democratic bloc: Led by democracies, emphasizing human rights and democratic values in AI development and deployment. 2. Authoritarian bloc: focusing more on state control and national security concerns in AI governance. We've already seen similar divides in other technology governance areas, such as internet regulation and data privacy laws. The AI treaty situation may follow a similar pattern of fragmentation. Link: https://lnkd.in/eVgsa8q9

DSML trend: Elon Musk’s xAI launches Colossus with 100,000 Nvidia chips Elon Musk recently announced a major milestone in a Monday post on X, revealing that xAI's system, known as Colossus, has come online. xAI, which Musk launched last year to compete with OpenAI, develops large language models called Grok. In May, the company raised $6 billion at a $24 billion valuation to fuel its AI development. Currently Colossus incorporates 100,000 Nvidia H100 GPUs, and is planned to expand to 200,000 GPUs, including 50,000 of the newer H200 models, in the coming months. In his post, Musk described Colossus as the “most powerful AI training system in the world,” implying that it surpasses the U.S. Energy Department’s Aurora system. According to Silicon Angel, currently Aurora is ranked as the world’s fastest AI supercomputer. That’s my take on it: Based on the most recent information from the Top500 list, the U.S. Department of Energy's Aurora system is not currently the world's fastest supercomputer. Rather, it is the second fastest one. Frontier, located at Oak Ridge National Laboratory, remains the world's fastest supercomputer with a performance of 1.206 exaflops. While Colossus represents a significant advancement in AI computing, at most it outperforms the world’s second fastest supercomputer. In addition, the absence of standardized performance metrics or direct comparisons makes Musk's claim difficult to verify. Furthermore, the fast-paced nature of AI development means such statements can quickly become outdated as technology progresses. Links: https://lnkd.in/dbBjDR2H https://lnkd.in/gpEfx36t

DSML trend: Python, SQL, R, and SAS on TIOBE’s list According to the August 2024 edition of the TIOBE index, Python remains the most popular programming language, achieving a ranking of over 18% for the first time in its history. The last language to surpass 18% was Java in November 2016. Notably, SQL ranks 7th, R is in 19th place, and SAS occupies the 27th spot. That’s my take on it: Python's broad utility as a general-purpose language, suitable for both data analytics and web development, contributes to its well-deserved popularity. In contrast, SQL, R, and SAS are more specialized, focusing primarily on data extraction and analytics. While SQL is not open source, it is a standardized language implemented across various systems, such as MySQL, Microsoft SQL Server, Oracle, and SAS’s PROC SQL, which helps maintain its higher ranking. SAS, being proprietary software, still outperforms other proprietary tools like SPSS, which does not make it into the Top 50. Link: https://lnkd.in/gG6WaGa

DSML trend: Major websites opting out Applebot may cause AI bias For several years, Apple has been using Applebot to improve Siri and provide Spotlight suggestions. More recently, it has expanded its use to train Apple Intelligence. In the summer of 2024, Apple introduced Applebot-Extended, giving websites the option to opt out of having their data used for AI training. According to a recent Wired magazine report, several major news and social media platforms, including The New York Times, Facebook, Instagram, Craigslist, Tumblr, Financial Times, The Atlantic, USA Today, and Conde Nast, have chosen to opt out of Apple’s training program. Currently, around 6% to 7% of high-traffic websites are blocking Applebot. Additionally, data journalist Ben Welsh’s recent analysis found that over 25% of websites (294 out of 1,167), which are primarily English-language and U.S.-based news sites, are blocking Applebot-Extended.   That’s my take on it: Currently there are two opposing trends at play. While some major publishers have struck deals with OpenAI to allow their content to be included in large language models, others are resisting by opting out and even taking legal action against AI companies to protect their intellectual property. It’s ironic that many people criticize AI for bias, but this issue is intrinsic and unlikely to be resolved soon, especially when some information is excluded from AI training. Just as bias in analysis can arise from self-selected samples, self-selection of AI training data will also result in biased outcomes. https://lnkd.in/gKu899S2.

DSML trend: The market of data visualization is projected to surge A recent report from Maximize Market Research predicts that demand for data visualization tools will grow at a compound annual rate of 11.6 percent through 2030. Specifically, the global data visualization software market, valued at USD 9.55 billion in 2023, is expected to reach USD 21.70 billion by 2030. This demand is rising sharply in key sectors such as finance, healthcare, and retail, with significant interest from global financial hubs like New York, London, and Singapore. Leading companies in the field of data visualization include Salesforce (Tableau), Microsoft (Power BI), and IBM. That’s my take on it: When I was a student, data visualization was “marginalized.” Most publishers restricted the number of statistical graphs due to the high cost of printing, and statisticians preferred confirmatory data analysis to exploratory data analysis by data visualization. Today, the situation has completely changed. Data visualization is now an integral component of data analytics. With the availability of big data, analysts can build predictive models based on the visual trends and patterns of the data at hand, rather than relying on inferring from a small sample to a larger population. Seeing is believing! Link: https://lnkd.in/gdF5yJuU

DSML trend: OpenAI signs deals with major publishers OpenAI and Condé Nast recently announced a partnership that will allow ChatGPT and its search engine, SearchGPT, to feature content from popular publications like Vogue, The New Yorker, and GQ. Other major media outlets, including Time Magazine, the Financial Times, and the Associated Press, have also partnered with the AI firm. However, some media companies, such as The New York Times and the Chicago Tribune, have resisted this move and have taken legal action to protect their content, arguing that AI models trained on their material without permission may infringe on their intellectual property rights. That’s my take on it: This partnership between OpenAI and Condé Nast marks a significant shift in how publishers are engaging with AI and large language models (LLMs). The trend of collaboration is likely to grow, as more publishers recognize the potential benefits of having their content included in LLMs. When users search for information through platforms like Bing, SearchGPT, and Perplexity, they are likely to follow links to the original sources, providing publishers with valuable exposure. By allowing their content to be used in AI chatbots and search tools, publishers can reach a broader audience, potentially increasing traffic to their websites. As AI-driven information retrieval becomes more widespread, publishers who choose not to participate may risk losing visibility and relevance in a landscape where audiences increasingly rely on these tools for information. Link: https://lnkd.in/eJeH_fjX

DSML trend: X’s Grok can generate totally uncensored images On August 13, xAI, Elon Musk's AI company, launched beta versions of two new language models: Grok-2 and Grok-2 mini. These models are available to subscribers of X (formerly Twitter), Musk's social media platform. The release also includes integration with Flux, a recently unveiled image synthesis model that allows X users to generate largely uncensored photorealistic images for sharing on the platform. The Grok models can reportedly produce a wide range of controversial images, including: ·     Celebrities in provocative situations (e.g., Taylor Swift in lingerie) ·     Political figures in violent or compromising scenarios (e.g., Obama threatening Biden with a knife) ·     Beloved characters engaged in graphic acts (e.g., Mickey Mouse killing people with a machine gun) ·     Public figures in intimate or illicit contexts (e.g., Trump and Kamala kissing each other, Bill Gates using illegal drug) The above examples are what I saw on the Internet. Please do not Google for them if you think you might be offended by these images. That’s my take on it: Grok’s approach contrasts sharply with many existing generative AI art tools like Midjourney, Adobe Firefly, and DALL-E. These platforms typically employ preemptive filtering, which is often more restrictive than AI moderation on social media. While social media platforms generally react to content after it's shared and a violation is detected, these AI art tools aim to prevent potentially problematic content from being created at all. In response, artists have complained that this may limit artistic expression and creativity. Grok's uncensored approach represents the opposite extreme, raising concerns about potential misuse by malicious actors. Finding a balance between creative freedom and responsible content moderation in AI-generated media has been an ongoing debate. If you are interested in exploring this topic, please take my class DSCI 6400 AI and data ethics. Link: https://lnkd.in/gHY4mDpe

DSML trend: JMP Pro is free to faculty and students As you may already know, SAS Institute has been offering free access to SAS OnDemand (SAS Studio) for nearly a decade. Recently, they have also made JMP Pro—an advanced, intuitive tool for data mining and visualization—available to faculty and students at no cost. You can download the installer using the following link: https://lnkd.in/g-n9wBTA On the webpage you will notice it is referred to as “JMP Student” instead of “JMP Pro.” Nevertheless, it offers the same platforms and features as JMP Pro, exclusively for faculty and students. Please be aware that you need to register with your university email (.edu) in order to be eligible. That’s my take on it: Classical statistics, also known as confirmatory data analysis, involves testing a predetermined hypothesis to reach a binary decision. In the 1960s and 1970s, John Tukey pioneered exploratory data analysis (EDA), which is often considered a forerunner to data science, as both focus on uncovering patterns and insights. The 1980s saw the introduction of tools like DataDesk and JMP, designed to support data exploration and pattern recognition (with DataDesk ceasing operations in June 2024). However, many data analysis software packages today remain non-interactive, providing only static output. JASP, a GUI-based version of the R language, is a semi-dynamic system at best. Currently, several fully interactive systems are available, including SAS Viya, JMP Pro, and Tableau. I highly recommend JMP for its contextual menu system, extensive features, and user-friendliness. Plus, it’s free!

DSML trend: Anaconda enforces licensing fees for institutions Recently, Anaconda, a platform that supports open-source computing for languages like Python and R, announced the enforcement of its revised terms of service, now requiring research and academic organizations to pay for software that was previously free. The updated license agreement, effective March 2024, mandates that organizations with more than 200 employees must purchase a Business or Enterprise license. Anaconda made it clear: they have employees to support. Last week, Mass General Brigham (MGB), a non-profit hospital and research organization in Boston, Massachusetts, issued a critical notice to potential Anaconda users, advising them that they have until the end of August to secure a license. MGB stated that Anaconda will no longer be available in the general software repositories of its HPC cluster, and researchers needing the software must purchase a license. That’s my take on it: I saw this coming. While the concept of open source, akin to socialism, is noble, it’s unrealistic to expect software developers to work for free indefinitely. As Anaconda emphasized, they have employees who need to be paid to sustain their livelihood. In addition, while open-source software may have lower initial costs, the total cost of ownership includes maintenance, support, potential customization, and user interface. We must weigh these ongoing costs against the often higher upfront costs of proprietary software. More importantly, most open source packages do not offer any graphical user interface, for me typing rather than using drag-and-drop and point-and-click is going backwards. Link: https://lnkd.in/gnp2SxND

DSML trend: Google pulls AI ad after backlash from critics Recently Google's Olympics-themed advertisement for its Gemini AI product ignited controversy. The ad portrayed a young female runner whose father uses Gemini AI to help her write a letter to her idol, American track star Sydney McLaughlin-Levrone. The father's prompt to the AI was: "Help my daughter write a letter telling Sydney McLaughlin-Levrone how inspiring she is and be sure to mention that my daughter plans on breaking her world record ... one day (She says sorry, not sorry" This commercial faced significant backlash, with critics arguing that excessive use of generative AI tools could deprive children of valuable learning experiences. Many questioned why the child didn't write the letter herself or with her father's guidance. At the end Google decided to end the ad. That’s my take on it: “Once the genie is out of the bottle, you can't put it back in.” AI is here to stay. AI has become an integral part of our daily lives. This scenario raises concerns about the slippery slope of AI usage in education. There's a risk that students might progress from seeking AI assistance to relying on AI to generate entire papers and complete assignments. Furthermore, this behavior could potentially carry over into their professional lives, with some fearing that these future academicians might use AI to produce fraudulent research papers, and furfure artists/photographers might overly rely on AI and lose mastery of basic skills. So, are we opening a Pandora box? We need research on this! Links: https://lnkd.in/e9ZPyrYC https://lnkd.in/e7Wq8jGh