AI: Humanity's Savior or Ultimate Destroyer?

WELCOME

Someone asked me yesterday where I think the “AI revolution” is going.

I honestly think artificial intelligence has the potential to become either the one thing that saves mankind or the curse that destroys humanity.

Because face it, there seem to be existential threats to humanity. Even though there is always war and disease, it seems like things are intensifying. Humans seem to be more in strife with each other than they are bonding and connecting.

I think AI might bring us solutions to problems that have become too big for us to solve collectively.

But AI might easily and utterly destroy humankind. We know what those scenario’s look like, and they’re not infathomable at all.

We ARE destroying the planet, and there are too many of us. Those are facts.

But the more I’m observing AI evolutions, the more I notice there’s a lot of smoke and mirrors. For example, while ChatGPT can be incredibly useful, when you use it a lot, you start to feel like it’s a party trick.

You cannot blindly rely on it, and at times it’s frankly a big, giant mess.

Claims are made but it's more about getting funding than anything else.

In the short term, I think nothing much happens. I’m waiting for the tipping point—where AI stops being the cool party trick and shows a spark of superintelligence of a quality that exceeds anything a human is capable of.

Until then, we keep our eyes focused on the latest AI news with the intensity of an eagle that hasn’t eaten in 2 weeks.

What do you think?

In any case, if humankind is ever in trouble, you’ll read it here first.

AI NEWS
Oil States Are The Solution For Chip Hungry AI

• Artificial intelligence and its development are hungry for computing power. A couple of events, like the Evergreen freighter getting stuck in the Suez Canal and obviously COVID-19, created a limited supply of semiconductors while demand was surging.
  
  Sam Altman, OpenAI’s CEO, is reportedly in talks with wealthy investment parties from the Middle East to set up their own chip manufacturing plants throughout the world. OpenAI has become largely dependent on Nvidia for the supply of computing power.
  
  OpenAI wants to make sure they’re able to keep feeding an animal that is getting increasingly hungry for "compute.”.
  

• Elon Musk has the same idea - and confirmed in an interview with Peter Diamandis that he’s building a 500 Megawatt GPU cluster in Kuwait. Interesting to see how “Mega-Oil” is pouring its vast wealth into AI development.

• An internal document leaked from Google revealed the tech giant’s ambitions in the field of AI. It literally states the goal is to “deliver the world’s most advanced, safe, and responsible AI.”. This statement seems extremely ambitious and disconnected from the current reality. Its models are lagging behind those of their competitors, Google Bard is yet to gain traction, and more importantly, Google has not found an answer to AI-generated spam poisoning their search engine results. It looks like Google is having an existential struggle.

• The strategy became apparent. Only a day after the aforementioned leak, Google Chrome will soon be updated with some new AI features. It looks like the era of the AI browser has officially begun.
  
  There’s a feature called “Tab Organizer” that automatically sorts open tabs into suggested groups, making it easier to manage multiple tasks. “AI themes” will allow users to personalize their browsers by selecting a subject, mood, and color for the AI to render.
  Next month, we’ll get a built-in “Help me write” feature to help you compose text on websites.
  

• A new study from MIT (link to PDF version of the paper) suggests that AI is unlikely to claim and replace as many human jobs as fast as initially projected. The gist of it is that building custom systems is still too expensive compared to just using human labor. The pace of AI cost reduction and performance improvement will largely impact the scale of disruption in job markets.

• Adobe announced they have developed an AI model called “ActAnywhere”.

You can take a picture of someone running - a still background, and the AI model merges the two together. Exactly what the world needs—more tools to create fake content. Way to go Adobe!
Just kidding - ’s impressive.

CYBERSECURITY NEWS
Cozy Bear Breaches Microsoft’s Email System

• Quite the week on the cybersecurity news front. Russian hacker group APT29 (aka “Cozy Bear”), known for their sophisticated cyber attacks, recently breached Microsoft's corporate email system, impacting the accounts of top executives and cybersecurity team members. The attack, revealed in a regulatory filing, used a technique called password spraying to gain access, but Microsoft assures there's no evidence of significant damage to their operations or customer data. The company is actively investigating the extent of the breach, working with law enforcement, and has taken steps to prevent future incidents.
  

• Just days after a critical vulnerability in Atlassian Confluence was revealed, cybersecurity experts have detected widespread attempts to exploit the flaw, known as CVE-2023-22527. The Shadowserver Foundation observed around 40,000 exploitation attempts from numerous IP addresses, primarily probing for vulnerable servers. Although 11,000 Confluence instances are exposed online, it's uncertain how many are susceptible to this specific attack, which primarily targets older versions of the software.

• Since late 2021, a Chinese hacking group known as UNC3886 has been exploiting a critical vulnerability in VMware's vCenter Server, identified as CVE-2023-34048. The flaw, which was patched in October, allowed the attackers to breach servers and deploy sophisticated backdoors for espionage purposes, primarily targeting defense, government, telecom, and technology sectors. Security firm Mandiant linked this activity to the group after observing related server crashes and noted the removal of crash logs, indicating deliberate efforts by the hackers to conceal their tracks.
  

• CISA has released an emergency directive to urgently patch Ivanti’s Connect Secure and Policy Secure. These VPN solutions are under active exploitation by nation-state actors.

PODCAST TIP OF THE WEEK
The Rest Is History

I wanted to switch it up a bit this week and give some non-tech recommendations.

I studied history at university and then lost track a bit, learning cybersecurity topics rather than delving into the past.

This exquisite podcast was able to rekindle my passion for history. It’s brought to you by Tom Holland - author of many books on Roman history, and Dominic Sandbrook , an expert on modern British and American history.

The topics are very wide and varied , deeply researched, and the conversation is accessible yet intellectual in soothing tones.

Sometimes they have some guests on, and in a delightful UK/Belgian crossover, I warmly recommend the one where they have Bart Van Loo on to talk about “Burgundy” ( De Bourgondiërs).

This entire podcast has been described as a true candy store for history buffs.

BOOK TIP OF THE WEEK
A Gentleman in Moscow

Usually, these book recommendations will be obviously non-fiction and on “hard” topics like cybersecurity, AI, and tech.

But I’d like to make one exception.

I’ve read hundreds of books in my life, and if you were to ask me what the very best book I’ve ever read is, this one is it.

We’re in 1922. The story centers around Count Rostov, a Russian noble who has received house arrest. And his house is the Metropol Hotel, right across the Kremlin in Moscow.

The story plays out within the confines of the Metropol.

It’s simply dazzling, and the only time I have ever wept knowing a story was to come to an end. Yes, I cried at the end. I’m man enough to admit this :)

Buy the book here : https://amzn.to/3vR5960

They’re about to waste the book and are turning it into a TV show featuring Ewan McGregor as Count Rostov. Before that happens, read this book.

AI PRODUCTIVITY TIP OF THE WEEK
Superhuman

If you’re struggling with getting your inbox down to zero then “Superhuman” might be something for you.

It’s a subtle combination of a tool that uses keyboard shortcuts to quickly triage incoming mail and AI tools. Especially if you write and work in English, it might be interesting.

Another feature is the use of "snippets,” which are bits of text that you frequently write in emails and can use to quickly reply to your correspondents.

The interface is quite slick and a breath of fresh air if you’re used to Gmail, for example.

Now, I’m not sure how this tool will evolve with Microsoft Copilot and Google Bard on the scene, but they haven’t catched up yet at the moment.

Check it out here

Closing Thoughts

That’s another one in the books. I’ve received some emails last week of readers really liking this new format, so I guess we’ll be keeping it.

There’s also something like an “Open Rate” for newsletters and last week we scored an amazing 65% open rate, which is spectacular. The average open rate for a tech newsletter is around 40%.

I am extremely grateful for you, the reader. THANK YOU for being here and see you next week !

And if you like it - please send it to a friend or colleague. Let’s keep this train moving forward !