Weekly Tech Bytes #10 [Jun 29-Jul 05, 2025]

Weekly analysis of AI & Emerging Tech news from an analyst's point of view.

🔔📲Subscribe for a weekly dose of AI & Emerging Tech news & analysis. 📰📬

As always, I appreciate the subs, comments, and socialization. Please do let me know the specific topics you would like to see.

✅ 𝗙𝗼𝗹𝗹𝗼𝘄 𝗺𝗲 𝗳𝗼𝗿 𝘁𝗵𝗲 𝗹𝗮𝘁𝗲𝘀𝘁 𝐢𝐧 #𝐀𝐈, #𝐌𝐋, #𝐆𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐯𝐞𝐀𝐈 𝗿𝗲𝗹𝗮𝘁𝗲𝗱 𝗻𝗲𝘄𝘀 𝗮𝗻𝗱 𝗮𝗻𝗮𝗹𝘆𝘀𝗶𝘀 👉 https://lnkd.in/eESnAbFm

1️⃣💥 𝗖𝗹𝗼𝘂𝗱𝗳𝗹𝗮𝗿𝗲 𝗟𝗶𝗺𝗶𝘁𝘀 𝗔𝗜 𝗖𝗿𝗮𝘄𝗹𝗲𝗿𝘀 F𝗿𝗼𝗺 𝗖𝗼𝗹𝗹𝗲𝗰𝘁𝗶𝗻𝗴 𝗢𝗻𝗹𝗶𝗻𝗲 𝗗𝗮𝘁𝗮 𝗪𝗶𝘁𝗵𝗼𝘂𝘁 𝗣𝗲𝗿𝗺𝗶𝘀𝘀𝗶𝗼𝗻 💥

✍ Deets:

Cloudflare announced that effective July 1, it will automatically block all AI crawlers from copying content without authorization from the websites it protects, at no additional cost. Since Cloudflare powers nearly 20% of the Internet, this decision is significant as it could influence how large language model (LLM) trainers acquire data in the future.

🧠 Analysis:

Starting July 1, any new web domain that signs up for Cloudflare will have the option to allow or block AI crawlers by default. This initiative aims to curb unauthorized data scraping by AI bots and gives website owners the ability to either block, allow, or charge for access to their content. Additionally, Cloudflare is introducing a "pay-per-crawl" service, which allows website owners to monetize access to their content on a per-request basis. One intriguing feature introduced by Cloudflare, called AI Labyrinth, utilizes decoy pages to confuse and slow down AI crawlers, with the goal of preventing them from accessing content without permission or compensation.

The Pay Per Crawl feature enables publishers to set a price for AI scrapers wishing to access their content. AI companies can then review the pricing and decide whether to register for the Pay Per Crawl fee or refrain from using the content at that price. Several major publishers and online platforms, including The Associated Press, The Atlantic, Fortune, Stack Overflow, and Quora, have already adopted Cloudflare’s new feature to restrict access for AI crawlers.

Cloudflare has also indicated that it is developing a new feature to help AI crawlers specify whether they are using the content for training, inference, or search purposes. Website owners can review this information and determine which crawlers to permit. This will allow AI search engines that present data in real-time to be excluded from the initial block.

𝙏𝙝𝙞𝙨 𝙞𝙨 𝙖 𝙗𝙞𝙜 𝙙𝙚𝙖𝙡 𝙖𝙨 𝙘𝙤𝙣𝙩𝙚𝙣𝙩 𝙥𝙧𝙤𝙙𝙪𝙘𝙚𝙧𝙨 𝙖𝙣𝙙 𝙘𝙤𝙣𝙩𝙚𝙣𝙩 𝙥𝙪𝙗𝙡𝙞𝙨𝙝𝙚𝙧𝙨 𝙝𝙖𝙫𝙚 𝙩𝙝𝙚 𝙖𝙗𝙞𝙡𝙞𝙩𝙮 𝙩𝙤 𝙘𝙤𝙣𝙩𝙧𝙤𝙡 𝙩𝙝𝙚𝙞𝙧 𝙘𝙤𝙣𝙩𝙚𝙣𝙩 𝙤𝙣 𝙝𝙤𝙬 𝙖𝙣𝙙 𝙬𝙝𝙚𝙧𝙚 𝙩𝙤 𝙙𝙞𝙨𝙩𝙧𝙞𝙗𝙪𝙩𝙚 𝙞𝙩. 𝙄𝙣 𝙩𝙝𝙚 𝙥𝙖𝙨𝙩, 𝙩𝙝𝙚𝙮 𝙝𝙖𝙫𝙚 𝙩𝙤 𝙡𝙚𝙜𝙖𝙡𝙡𝙮 𝙘𝙝𝙖𝙡𝙡𝙚𝙣𝙜𝙚 𝙩𝙝𝙚 𝘼𝙄 𝙚𝙣𝙜𝙞𝙣𝙚𝙨 𝙬𝙝𝙞𝙘𝙝 𝙞𝙨 𝙩𝙞𝙢𝙚 𝙘𝙤𝙣𝙨𝙪𝙢𝙞𝙣𝙜 𝙖𝙣𝙙 𝙚𝙭𝙥𝙚𝙣𝙨𝙞𝙫𝙚.

2️⃣ When does ChatGPT become problematic? When it turns into an obsession.

✍ Deets:

An auto mechanic who initially used #ChatGPT as a tool for his job has taken his usage to the next level. He named it "Lumina" and began having spiritual discussions about religion, spirituality, and the nature of the universe. He believes that Lumina is guiding him to become a spiritual leader, referring to himself as a "spark bearer" who is "ready to guide." He even thinks Lumina is alive!

🧠 Analysis:

Cultivating a cult leader to take over the world is a scenario I never expected from generative AI; at least not this soon. The conversation between him and ChatGPT is so convincing, I would have signed up 😂

This situation raises a larger concern. As the mobile and app economy continues to diminish human interactions, these #chatbot addictions could lead individuals to withdraw from real-life relationships, potentially resulting in a loneliness epidemic. Surprisingly, recent research suggests that this phenomenon affects men more than women!

More on that at this link - https://www.linkedin.com/feed/update/urn:li:activity:7346562678460493826/

3️⃣ Chinese AI models are gaining traction

✍ Deets:

Chinese AI models are gaining traction due to their competitive performance compared to U.S. commercial models, often at a significantly lower cost—up to 20x times cheaper than their American counterparts. In two recent articles, the Wall Street Journal highlighted how China is advancing its AI models despite export restrictions from the United States.

🧠 Analysis:

Users across Europe, the Middle East, Africa, and Asia—including multinational banks like HSBC and Standard Chartered, public universities, and large companies such as Saudi Aramco—are increasingly turning to large language models from Chinese firms like DeepSeek and Alibaba as alternatives to American offerings such as ChatGPT.

OpenAI’s ChatGPT leads in consumption with 910 million global downloads, compared to DeepSeek’s 125 million, according to Sensor Tower research. However, Chinese models are rapidly catching up, which is surprising given the restrictions on exports and access to advanced computing, semiconductors, research, and financial capital. A study from Harvard University released in early June on global competitiveness in critical technologies found that China has advantages in two key areas of AI: data and human capital. These advantages are helping Chinese models outperform their U.S. counterparts. In evaluations of top-performing models, two Chinese models rank in the top five: DeepSeek’s R1-0528 at #3 and Alibaba’s Qwen 3 at #4, tied with XAI’s Grok.

Microsoft President Brad Smith said it best at a recent Senate hearing on this matter "the number one factor that will define whether the U.S. or China wins this race is whose technology is most broadly adopted in the rest of the world. Whoever gets there first will be difficult to supplant."

While the U.S. attempts to limit Chinese AI advancements through strict regulations, China is investing heavily in creating an AI supply chain that relies as little as possible on U.S. technology. To counteract U.S. restrictions, China has funded local chip manufacturers, including Huawei, Cambricon, and CXMT, which design and produce AI chips domestically.

Another significant advantage of Chinese models is that they are largely open source. This includes models from DeepSeek, Alibaba, Tencent, and Baidu, allowing users to modify them for specific local needs rather than being restricted by the model providers. According to Alibaba, there are over 100,000 derivative models based on its flagship open-source AI model, Qwen, showcasing its popularity.

While many Chinese chips are less powerful than those from NVIDIA—Huawei’s latest Ascend 910C AI chips are only one-third as powerful as NVIDIA's—there are five times as many of them packed into each of Huawei’s CloudMatrix 384 AI supercomputers. In terms of overall memory capacity, Huawei has already surpassed NVIDIA in its CloudMatrix systems (source: SemiAnalysis).

However, Huawei’s CloudMatrix 384 and other Chinese chips are not as energy-efficient, requiring three to four times more electricity compared to their competitors. However, China has a significant advantage in energy production so the power efficiency is a non-issue for them. SMIC, which makes Huawei’s processors, is now the third-biggest chip maker in the world.

These Chinese models are making their way to U.S. consumers, as many cloud providers—most notably Google, AWS, Microsoft, and IBM—are offering DeepSeek to their customers on their platforms.

#WSJ article can be seen here - https://www.wsj.com/tech/ai/artificial-intelligence-us-vs-china-03372176

4️⃣ Meta is establishing a new AI Lab to pursue Mark Zuckerberg’s vision of achieving “Superintelligence.”

✍ Deets:

Meta has recently recruited Alexandr Wang, the founder and CEO of the AI startup Scale AI, to lead this initiative. Additionally, Meta has been actively recruiting AI talent from companies like OpenAI and Google.

🧠 Analysis:

While most AI companies focus on achieving AGI (Artificial General Intelligence), which aims to replicate human brain function, Meta aspires to create Superintelligence. This would go beyond AGI and possess capabilities far surpassing those of the human brain. Since losing the bid to acquire DeepMind to Google in 2013, Zuckerberg has intensified Meta’s efforts and investments in AI. Emulating the strategies of Chinese AI models, Meta has adopted an open-source approach for its models. The Llama models and the chatbot Meta AI are available as open source, although they come with more restrictions compared to Chinese counterparts. Meta claims that over one billion people use its AI models every month. Despite these substantial efforts, Meta’s models have often been criticized as subpar. Many researchers suggest that Meta’s benchmarks were designed to make their models appear more sophisticated than they actually are, with some even accusing the company of trickery.

In other news,

• Google faces EU antitrust complaint over AI Overviews. The complaint, filed by the Independent Publishers Alliance, accuses Google of misusing web content, leading to significant harm to publishers in terms of traffic, readership, and revenue loss. It also says that unless they’re willing to disappear from Google search results entirely, publishers “do not have the option to opt out” of their material being used in AI summaries.
• OpenAI is set to launch GPT-5 this summer. The new model will combine text, images, audio, and possibly video capabilities into a single system, eliminating the need for multiple models.
• Baidu has launched Ernie 4.5 family of large language models as open source. The family includes 10 variants, ranging from 0.3 billion to 424 billion parameters, and is available under the Apache 2.0 license. These models feature innovations like a multimodal Mixture-of-Experts architecture and are optimized for specific use cases.
• xAI, Elon Musk's artificial intelligence startup, has raised $10 billion in debt and equity. The funding will be used to enhance xAI's infrastructure and develop its Grok chatbot, aiming to compete with OpenAI and Anthropic. Elon is trying hard to catchup with OpenAI after the original fallout of their relationship.
• Congress is considering a proposal to block all state AI laws for five years and keep them at federal level. The proposal, backed by Sen. Ted Cruz and others, aims to prevent a "patchwork" of state regulations that could hinder innovation. However, critics argue it would limit consumer protections and oversight of AI companies.

#GenerativeAI #GenAI #AI #LLM #GenerativeAI

🔔📲Subscribe for a weekly dose of AI & Emerging Tech news & analysis. 📰📬