• AI KATANA
  • Posts
  • ChatGPT website traffic has fallen for the first time

ChatGPT website traffic has fallen for the first time

Also: Sarah Silverman is suing OpenAI and Meta for copyright infringement

Author

AI KATANA
July 10, 2023

Hey!

Today we explore a noteworthy dip in ChatGPT's website traffic for the first time since its launch - a trend that incites debate on the lasting appeal of AI chat tools. In legislative updates, we assess Switzerland's cautious stance on AI regulation, as it navigates the intricate balance of fostering innovation while addressing ethical considerations. In legal developments, OpenAI and Meta face copyright infringement lawsuits from Sarah Silverman and authors Christopher Golden and Richard Kadrey, which may have far-reaching implications for AI companies. Turning to Southeast Asia, we examine Indonesia's unique position to influence AI governance within the region, supported by a strong digital economy and academic prowess. In AI tools and VC news, Microsoft's LongNet opens new doors for sequence length scaling, while AI startups continue to defy the downward trend in VC funding, reflecting the enduring fascination with AI potential.

Sliced:

  • 📉 ChatGPT website traffic has fallen for the first time

  • 🇨🇭 Where does Switzerland stand on regulating AI?

  • 👩🏻‍⚖️ Sarah Silverman is suing OpenAI and Meta for copyright infringement

  • 🇮🇩 Indonesia’s chance to influence AI governance in Southeast Asia

For the first time since its launch, OpenAI's ChatGPT has experienced a dip in website traffic, with a 9.7% fall in global traffic in June, a 5.7% drop in unique visitors, and an 8.5% decrease in the time spent by visitors on the site, according to Similarweb. Despite the decline, ChatGPT's site still attracts more visitors than Microsoft's search engine Bing and the second most popular standalone AI chatbot site, Character.AI. Some suggest the fall in traffic signals a fading novelty in AI chat, but this conclusion may be hasty as the data only considers website traffic, not usage through newer mobile apps or Microsoft's incorporation of ChatGPT elements into Bing. Moreover, the rise of ChatGPT has increased interest in AI tools in general, including text, image, and video generation software, potentially shifting user exploration towards these other tools.

Switzerland, like other countries, is grappling with the need for artificial intelligence (AI) regulation but is currently adopting a "wait and see" approach as they analyze their options. The country is taking note of the progress made by the European Union, which has presented a draft for AI legislation, and the Council of Europe's proposed framework. However, Swiss representatives believe that there won't be a single overarching law for AI, drawing parallels with engines, where regulations differ based on specific uses. Existing laws, from the constitution to the protection of fundamental rights, cover some aspects, but AI presents new challenges that require bridging these gaps. When AI regulations are established, questions of responsibility and liability will be significant, with the belief that human actors must be identifiable as they remain responsible for the development and application of AI systems.

Comedian Sarah Silverman and authors Christopher Golden and Richard Kadrey are suing OpenAI and Meta for alleged copyright infringement, claiming their AI models were trained on datasets containing their works without consent. The trio asserts OpenAI's ChatGPT and Meta's LLaMA were trained using datasets sourced from "shadow library" websites, including their works available in bulk via torrent systems. Demonstrative exhibits showcase ChatGPT summarising their books, infringing their copyrights. Similar claims against Meta pertain to their LLaMA models. The plaintiffs are seeking statutory damages, restitution of profits and more for six counts of copyright violations, negligence, unjust enrichment, and unfair competition. This litigation challenges the boundaries of copyright, signalling potential repercussions for other AI companies.

Indonesia has a great opportunity to become a leader in AI governance in Southeast Asia. The country has a strong track record of applying AI in both the public and private sectors, and it is also the region's largest potential market for AI development investment. If Indonesia can continue to develop its AI capabilities and establish strong governance frameworks, it could play a major role in shaping the future of AI in Southeast Asia. The article highlights several factors that make Indonesia well-positioned to lead in AI governance. First, the country has a large and growing population, which provides a large pool of potential users and developers for AI-based products and services. Second, Indonesia has a strong government that is committed to developing the country's digital economy. Third, Indonesia has a number of universities and research institutions that are conducting cutting-edge research in AI.

🛠️ AI tools updates

LongNet, a novel Transformer variant, addresses the demand for scaling sequence length in large language models, boasting the ability to scale to more than 1 billion tokens without compromising performance on shorter sequences. To accomplish this, LongNet introduces dilated attention, which exponentially expands the attentive field as the distance increases. The model benefits from linear computational complexity, a logarithmic dependency between tokens, the capacity to serve as a distributed trainer for extremely long sequences, and seamless integration with existing Transformer-based optimizations due to its drop-in replacement for standard attention. Test results suggest LongNet performs strongly in both long-sequence modeling and general language tasks, offering unprecedented opportunities to model extremely long sequences, such as an entire corpus or even the whole Internet.

💵 Venture Capital updates

Despite an overall decline in venture capital (VC) funding in recent years, AI startups have defied the trend, attracting $15.5 billion in 2023, exceeding the total raised last year and reaching over halfway to 2021's peak of $9.1 billion. This surge, including investments into early-stage startups like Mistral AI and Inflection, signals a persistent herd mentality among VCs who are keen not to miss out on potential AI breakthroughs. While critics argue that this trend risks inflating valuations, proponents believe the influx of capital could stimulate technological progress and sector growth. However, in the face of market uncertainty, investing off the beaten track could carry greater risks, particularly for less-experienced VCs. Despite the continued hype, some believe that today's investors are more discerning than in the past, seeking out category-defining businesses and focusing on innovative technology and solid metrics rather than blindly following the herd.

🫡 Meme of the day

⭐️ Generative AI image of the day