- The Blacklynx Brief
- Posts
- The AI Chip Wars
The AI Chip Wars
WELCOME
Two weeks ago one of our main news topics was the fact that Elon Musk’s Neuralink had planted the first chip into a human brain. On Tuesday this week Musk said that this person has recovered nicely and is now able to move a mouse over a computer screen just by thought alone.
I’m a bit freaked out by that and also slightly excited by the advances being made.
A notable trend we see emerging is that there is not only a race between different tech companies to build better AI algorithms, a more important battle is being waged over computing power and the capability to develop the microchips.
It’s a full scale war for computing power. Very fascinating to see that emerge.
Welcome to the Blacklynx Brief !
AI NEWS
The AI Chip Wars
We talked about this in a separate post on LinkedIn last Friday, but OpenAI introduced Sora, a text-to-video AI model capable of creating videos up to 60 seconds long from text and images, combining features of GPT and DALL-E for realistic results. It is currently available to select users for early feedback, with speculation that it was developed using synthetic data from Unreal Engine.
It cannot be understated how important this development is. This might spell the end of an entire industry.Recent analysis shows that AI is significantly impacting the freelance job market, leading to a decrease in writing and customer service roles, but a surge in demand for AI-related skills, video editing, graphic design, and web/software development. Specifically, chatbot development demand has increased by 2000% as companies prefer integrating existing AI APIs. This shift highlights the evolving job market due to AI advancements, with ongoing changes expected as AI technology and integration progress.
Elon Musk announced on X (formerly Twitter) potential AI developments including a partnership with Midjourney to possibly integrate its image generator into X’s Grok chatbot, plans for X to label AI-generated images which may align with a Midjourney collaboration, and the upcoming release of Grok 1.5 with new features. Additionally, Musk shared that the first human Neuralink patient has successfully recovered, achieving the ability to control a computer mouse through thought
Google unveiled the Gemini 1.5 model, enhancing its AI systems with a 1M token context window, potentially expanding up to 10M tokens, allowing for processing of significantly larger data sets including vast text, code, audio, and video. This upgrade provides the ability to delve into detailed analyses of substantial content, setting a new standard in data processing capabilities. Initially, Gemini 1.5 will be available with a 128K context window, with plans to scale up, currently restricted to select developers and enterprise users.
Masayoshi Son, founder of SoftBank, plans to raise $100B for 'Project Izanagi', an AI chip venture aiming to compete with NVIDIA and advance towards achieving AGI, leveraging significant funding from SoftBank and Middle Eastern investors. This initiative seeks to not only challenge the leading chipmaker but also complement Son's existing chip design firm, Arm Holdings, with a vision of realizing AGI within the next decade. Izanagi represents a major investment in the AI sector, indicating a strategic move to secure necessary resources amidst growing demand and potential shortages in the AI chip market.
Reddit has entered into a $60M annual agreement with Google for the use of its user-generated content for AI model training, ahead of its planned IPO. This deal has raised concerns among users about the commercial use of their content without compensation, especially as Reddit considered measures like blocking Google search crawlers to secure such an AI data deal. This agreement could serve as a precedent for future AI licensing deals, highlighting the growing value of social media data in AI development and the potential financial benefits for platforms like Reddit.
French AI startup Mistral has introduced a new model named 'next' for testing, accessible via LMSYS Direct Chat and subtly announced in a Discord comment. Initial feedback from users suggests its performance in coding and reasoning tasks is comparable to, or surpasses, GPT-4. Following a significant funding round of $415M in December, valuing Mistral at $2B, the company is emerging as a formidable competitor in the AI field. This development could signify a pivotal moment for open-source AI, particularly if 'next' proves to be superior to GPT-4.
AI startup Groq has made significant advancements with its specialized processing units, known as LPUs (language processing units), designed to run large language models (LLMs) with nearly instantaneous response times. These LPUs provide a more efficient alternative to traditional GPUs (graphics processing units), with public benchmarks showing Groq's hardware achieving speeds of 500 tokens per second, drastically outpacing the 30-50 tokens per second typical of GPT 3.5. Groq's technology, showcased in a demo, demonstrates a performance about four times faster than GPT 3.5. Founded in 2016, Groq has also legally challenged Elon Musk’s X for using the name 'Grok' for its chatbot. Groq's breakthroughs suggest a transformative potential for AI user experiences and present an efficient, affordable alternative to GPUs.
CYBERSECURITY NEWS
Lockbit ransomware gang is shut down
Law enforcement agencies from 11 countries collaborated in Operation Cronos to dismantle the LockBit ransomware operation, seizing its infrastructure, freezing cryptocurrency wallets, releasing decryption tools, and arresting members. The operation, led by the UK's National Crime Agency, involved infiltrating LockBit's servers, mapping its infrastructure, and accessing its backend. The takedown, highlighted by the replacement of LockBit's dark web portal with a seizure banner, was announced through official press releases from the UK NCA, Europol, and the US Department of Justice.
Microsoft, in partnership with OpenAI, reported detecting state-affiliated hacking groups from China, Russia, and Iran utilizing AI tools, particularly language models, to enhance their cyber operations. These groups, identified by code names such as Forest Blizzard, Emerald Sleet, Crimson Sandstorm, Charcoal Typhoon, and Salmon Typhoon, have been using large language models (LLMs) to augment their hacking capabilities. This insight presumably comes from monitoring chatbot interactions, suggesting that Microsoft and OpenAI analyze user interactions with AI tools to identify malicious activities, in line with their terms of service. This confirms the companies' monitoring of AI usage to detect and report on cybersecurity threats.
Japan has observed a significant increase in cyber incidents, especially from China-backed hackers targeting crucial sectors like telecom and internet services. The reported incidents rose from 150 in 2021 to 230 in 2022, with a continued upward trend. To combat these threats, Japan is enhancing collaboration with the U.S. and other countries, while domestically, the National Center of Incident Readiness and Strategy for Cybersecurity (NISC) has increased its resources. These efforts aim to bolster Japan's cybersecurity defenses and address the growing challenge of international cyber threats.
The Federal Trade Commission (FTC) is addressing the rise in AI-driven impersonation fraud by proposing an expansion of a new rule, initially designed to protect governments and businesses, to also cover individuals. This move comes in response to growing incidents of fraud involving advanced technologies like deepfakes. The proposed rule aims to empower the FTC to initiate federal court actions to reclaim funds from scams that exploit the identities of businesses and government entities. Additionally, the agency is considering making it illegal for businesses, particularly those in AI, to supply products or services that facilitate impersonation fraud. This reflects the FTC's efforts to curb the misuse of AI in executing sophisticated impersonation schemes that target consumers.
Norway's Prime Minister Jonas Gahr Støre highlighted at the Munich Security Conference that a conflict between Russia and European states could escalate beyond military confrontation, focusing on the threats posed by fake news, conspiracy theories, and cyberattacks on democracies. Amidst concerns about NATO's defense commitments and comments from former U.S. President Trump, Dutch Prime Minister Mark Rutte expressed skepticism regarding NATO's response under Trump's administration. Støre emphasized the importance of protecting democracies, especially considering recent espionage activities targeting Norway's critical oil and gas sectors, underlining the role of NATO's non-EU members in countering potential Russian aggression.
The European Union has initiated an investigation into TikTok for potential breaches of the Digital Services Act (DSA), focusing on child protection features and harmful app design practices. This inquiry, placing TikTok as the second company under scrutiny after Twitter, will also examine TikTok's adherence to ad transparency rules and its provision of access to research data. The DSA, adopted last year, aims to regulate large online platforms.
The Anatsa banking Trojan, known for targeting European users, has expanded its reach to include Slovenia, Slovakia, and Czechia, in addition to Germany, Spain, and the UK. It infiltrates devices through fake cleaner and PDF reader apps on the Google Play Store, with an estimated 150,000 downloads.
PODCAST TIP OF THE WEEK
Diary of a CEO
Steven Bartlett is an English entrepreneur and investor and also the newest member of the lauded TV show “Dragons Den” on BBC.
While only in his twenties he founded multiple companies and he took to podcasting in 2017. As of 2021 it was Europe’s most downloaded business podcast. Since then “Diary of a CEO” has become a worldwide phenomenon.
The production is top-notch - it seems it is filmed from the same location but he actually flies the set to the location of the guests.
He has some amazing guests and this will only get better as the popularity grows.
The only problem I have with the podcast is that Steven sometimes comes across as .. well… ignorant about the world. I have the impression he’s not well read. Sure he’s a genious in marketing, but he’s not able to challenge some of his guests.
So on a podcast with Jordan Peterson he comes across as a 6-year old taking to an adult.
He did not finish high school and it honestly shows in this format. But this was a recommendation after all , and I really like what he’s doing and where this podcast is going.
BOOK TIP OF THE WEEK
Adrian Tchaikovsky - Children of Time (2015)
Time for another fiction recommendation. If you’re into science fiction, this one is really not to be missed.
Earth is dying. Expeditions are sent out to find a new world and terraform it.
After that, the plan was to throw 10.000 monkeys on the new planet and release a special virus onto the planet that would accelerate these monkey’s evolution so they would evolve into humans by the time the refugees from Earth arrived.
Crazy premise - but it gets crazier.
Something goes horribly wrong in the release of the monkeys - but the nanovirus gets deployed anyway.
When the refugees arrive on the new planet they find they’re up against a race of bionic superspiders.
While it sounds like the premise of a Paul Verhoeven movie (Starship Troopers) - it’s actually quite smart and thoughtful with plenty of philosophical questions about what life is.
Closing Thoughts
That’s it for us this week.
Can I ask you a favor ?
If you’ve gotten to this part , it means you really read the entire thing.
This means you not only like pain, you like prolonged pain. Just kidding. It means you’re cool.
What would make you even more cool is that you send this link to a few people and tell them to subscribe.
Reply