• AI KATANA
  • Posts
  • Top AI Shops Fail Transparency Test

Top AI Shops Fail Transparency Test

Also: ScaleAI wants to be America’s AI arms dealer

Author

AI KATANA
October 22, 2023

Hi!

In the latest AI updates, Stanford's CRFM report found a significant transparency gap among top AI firms, with none meeting adequate standards. ScaleAI has aligned its offerings to aid the U.S. military, particularly after securing a contract with the Department of Defense and introducing a chatbot for the Army. IBM's NorthPole processor chip, inspired by the human brain, promises speedier AI operations but shows limitations for extensive language models. The luxury watch market is seeing a revolution with AI-driven tools enhancing authenticity and consumer engagement. Google's Search Generative Experience poses challenges for publishers, potentially impacting web traffic and revenue. On the tools front, YouTube's AI tool, allowing users to mimic famous musicians, has been postponed due to licensing hurdles, and an innovative AI tool for supernova detection has been introduced by researchers. Lastly, Chinese tech behemoths Alibaba and Tencent have invested in Zhipu AI, highlighting growing investments in China's AI sector.

Slliced:

  • ⚠️ Top AI Shops Fail Transparency Test

  • 🛡️ ScaleAI wants to be America’s AI arms dealer

  • 🤯 ‘Mind-blowing’ IBM chip speeds up AI

  • ⌚ How AI Is Shaping The Future Of The Luxury Watch Market

  • 🤖 As Google pushes deeper into AI, publishers see fresh challenges

A recent report by Stanford's Center for Research on Foundation Models (CRFM) has assessed the transparency of major AI companies in relation to their AI models. Despite previous commitments to transparency from 15 leading AI companies, the findings reveal a significant lack of clarity in the sector. Out of the ten foundation models evaluated on 100 different transparency metrics, Meta’s Llama 2 scored the highest with a mere 54 out of 100. The report indicates that none of the major foundation model developers offer adequate transparency. Key transparency metrics encompass training data sources, model properties, and post-release details. The study also highlights concerns regarding undisclosed labor practices, especially around the use of low-wage workers for model refinement. While some models have been praised for their transparency, debates continue over the wisdom of open sourcing such powerful tools due to potential misuse by malicious entities. The CRFM team plans to update this index annually, hoping it will guide global AI legislation.

ScaleAI, a tech startup founded by Wang in 2016, positions itself as a critical ally for the U.S. military in its competition against China's advancing technological prowess. Initially established to aid companies in data organization for AI training, Scale has since taken on a pronounced defense orientation. After securing a $249 million contract with the Department of Defense and partnerships with various military entities, in May, it became the first to deploy a large language model on a classified network through a collaboration with the Army's XVIII Airborne Corps. This chatbot, "Donovan", is designed to quickly summarize intelligence to aid commanders in decision-making. Wang, viewing the U.S. as lagging in the AI race against China, emphasizes the paramount importance of data in AI warfare. While Scale has thrived, even making Wang a billionaire, it faces stiff competition from tech giants like Google, Microsoft, and Amazon. Moreover, its dealings in the Global South have garnered criticism. The increasing militarization of AI technology has raised ethical concerns, particularly about the risk of diminishing human oversight in lethal decisions. Wang's commitment to the defense sector was strengthened after witnessing the intimate ties between China's tech industry and its military, fearing a potential AI military dominance by China over the U.S. Despite challenges in dealing with the defense sector, Wang envisions Scale's role as ensuring America's technological leadership in the rapidly evolving tech landscape.

IBM's NorthPole processor chip, a brain-inspired creation, significantly accelerates AI operations by minimizing external memory access, thus enhancing computation speed and energy efficiency notably in image recognition tasks. This innovation, bypassing the Von Neumann bottleneck, demonstrates a large-scale integration of computing and memory, hinting a paradigm shift in computer architecture. Although exemplary in certain aspects, NorthPole's design reveals limitations in handling extensive language models due to its RAM capacity, necessitating further advancements for broader applications including large-scale neural network processing.

AI and computer vision are modernizing the luxury watch market, notably enhancing authenticity verification and consumer engagement. AI's precision in analyzing minute details significantly mitigates the $5 billion counterfeit issue in the secondhand luxury watch sector, bolstering their value as alternative investments. Furthermore, live video commerce facilitates real-time, interactive buyer-seller interactions, overcoming traditional sales inefficiencies. Platforms like Wrist Shot, embedding AI subtly, are pivotal in nurturing trust between buyers and sellers amidst the growing pre-owned luxury watch market, projected to reach $30 billion by 2025. This digital transition, blending tradition with technology, signifies a promising evolution in luxury watch commerce.

Google's deeper venture into generative AI with its Search Generative Experience (SGE) has left publishers uneasy. SGE, which creates summaries for certain search queries, could deter users from clicking on the actual source links, potentially slashing publishers' web traffic and ad revenue. While Google introduced Google-Extended allowing publishers to block their content from being used to train its AI models, it doesn't extend to SGE. This move ignites fresh challenges in the long-standing tension between publishers and Google, as they navigate the evolving landscape of content discovery and monetization in the AI era.

🛠️ AI tools updates

YouTube is developing an artificial intelligence tool that enables users to replicate the voices of renowned musicians in their videos. Despite initial plans for a September launch during the "Made On YouTube" event, the introduction has been delayed due to ongoing licensing negotiations with major record labels, including Universal Music Group, Sony Music Entertainment, and Warner Music Group. While some music industry leaders see the potential of AI to enhance musical creativity, concerns arise over artist rights, potential misuse of the tool, and copyright issues, as exemplified by UMG's removal of a popular AI-generated track featuring the voices of Drake and The Weeknd.

Researchers have developed an AI tool, named Bright Transient Survey Bot (BTSbot), which autonomously identifies and confirms supernovas. Spearheaded by scientists at Northwestern University, this tool streamlines supernova studies, aiding a better understanding of stellar life cycles and the universe's origins. This innovation marks a significant shift from the traditional human-machine collaborative supernova discovery process, heralding a new era of astronomical research and freeing scientists for more in-depth analysis.

💵 Venture Capital updates

Chinese tech giants Alibaba and Tencent, along with other investors, have funneled US$342 million into Zhipu AI, a startup aspiring to rival Microsoft-backed OpenAI in generative AI. This funding surge, doubling Zhipu AI's valuation to US$1 billion since July, underscores the escalating investments in China's AI sector despite US restrictions. The capital will accelerate the development of Zhipu AI's large language models, propelling it closer to global competitors in creating advanced generative AI technologies, and fortifying China's standing in the global AI arena.

🫡 Meme of the day

⭐️ Generative AI image of the day