AI's Double-Edged Sword: Innovation, Intrusion, and the Quest for Ethical Governance

Welcome to Aideations. The most comprehensive daily AI newsletter on the planet! How do I know? Because I read over 50 of them, plus the news, so you don’t have to. It’s my goal to make this the most-read newsletter on AI. I can’t do that without your support and feedback.

Here's what we've got in store for you today:

🔌 ChatGPT Account Theft

👀 When Your Boss Becomes Big Brother

🦸 WEF Forms AI Alliance

💊 AI Is Revolutionizing The World Of Drug Discovery

📚 Research Of The Day

🎥 Video Of The Day

🛠 Tools Of The Day

🤌 Prompt Of The Day

🐥 Tweet Of The Day

Group-IB Exposes Massive Theft of ChatGPT Accounts on the Dark Web

Group-IB, a cybersecurity big shot stationed in Singapore, has discovered over a whopping 101,134 devices that are infested with stealer malware and contain saved ChatGPT credentials. These credentials were detected in logs of info-thieving malware traded on the shady corners of the dark web within the past year. If you had any doubts about your popularity, this should do the trick—the peak number of compromised accounts reached a dizzying 26,802 in May 2023 alone.

Let's put it this way: if you've been chatting up with ChatGPT, and you're in the Asia-Pacific region, there's a decent chance that some dark web dealmaker is currently trying to hawk your info. Seems like Asia-Pacific is the favorite shopping hub for these chatbot credentials, much like its reputation for cheap, quality electronics.

You're probably thinking, "Big deal, they've got my chat history with a bot." But consider this: lots of employees are using ChatGPT to streamline their work. So, that innocent chat you had with your AI buddy about the secret sauce in your company's product? Yep, that could now be in the hands of someone with a nefarious agenda.

So, how does Group-IB track all this? They've got their Threat Intelligence platform—the bibliothèque nationale of dark web data—that keeps tabs on the underworld of cybercrime forums, marketplaces, and hidden communities in real-time. The most common culprit behind the ChatGPT account breaches? A nasty piece of work called the Raccoon info stealer. Picture this character as a digital pickpocket with an insatiable appetite for data.

Now, let me give you the 101 on info stealers. These pieces of malware are digital con artists that swindle credentials saved in browsers, bank card details, crypto wallet information, cookies, browsing history, and more. These guys go for quantity over quality, infecting as many computers as they can. Their haul is then auctioned off in the dark web's equivalent of eBay.

After doing some digital detective work, Group-IB's Threat Intelligence unit found out that the Asia-Pacific region was the hot spot for ChatGPT account thefts. Between June 2022 and May 2023, info stealers swiped a jaw-dropping 40.5% of all ChatGPT accounts in the region.

Dmitry Shestakov, the Sherlock Holmes of cybersecurity at Group-IB, warns that as companies increasingly rely on ChatGPT, they're inadvertently handing over a treasure trove of sensitive intel to the dark side. Not exactly the kind of 'sharing is caring' we had in mind, eh?

So, what's a ChatGPT user to do? First off, change your passwords like you change your socks. Also, add an extra layer of security by enabling two-factor authentication. It's like a secret handshake only you and your account know.

Knowledge is power, and in this case, that power comes from having a window into the dark web. It allows organizations to nip threats in the bud, alert affected parties, and bolster their security to avoid future debacles. So, if you value your company's secrets (and your late-night chats with your AI buddy), it might be time to double down on your cybersecurity game.

When Your Boss Becomes Big Brother: The Rise of AI Surveillance at Work

In an unsettling plot twist straight out of a sci-fi thriller, your boss might now be playing spy with AI-powered surveillance tools. They're using algorithms and analytics, not to battle some international conspiracy, but to peek into your everyday work life. Companies like FiveCast, once peddling their wares to the military, are now offering their advanced tracking tools to corporate honchos. Your social media posts, your networking circles, even your emotions are now fair game.

These tools, despite being riddled with inaccuracies and vulnerabilities, are making inroads into workplaces. Even when they don't quite hit the mark, they still manage to create a culture of fear. The tag of being a "troublemaker" based on an algorithm's interpretation can have serious repercussions, not to mention the chilling effect it has on your legally protected right to protest or organize.

Defending their wares as merely "corporate awareness monitoring," these surveillance tech companies are thriving in the shadows. They argue that they only use public information, conveniently side-stepping the conversation on the potential misuse of these tools against a free society. Essentially, they're offering up the resources of a private CIA to any middle manager with a budget to spare.

In response to this creeping intrusion, transparency and regulation are needed. Companies should openly disclose their use of these tools, and we need new laws outlining what's acceptable in AI surveillance. This has to be a global initiative because multinational corporations can easily sidestep country-specific protections. Workers and unions should be key players in shaping these regulations.

Ultimately, there might be a need to challenge the very existence of this surveillance market. The trade in AI-generated profiles, peering into our emotions, associations, and thoughts, could be deemed illegal. Our personal life outside of work should be off-limits, free from the intrusive gaze of corporate interests. As we navigate this new landscape, it's crucial to question just how much privacy we're willing to surrender in the name of 'corporate security'.

The World Economic Forum's AI Governance Alliance

Ready or not, generative AI is kicking the door wide open and it’s proving to be a real double-edged sword. On one hand, it's got the potential to completely revolutionize how we work and create, with automation and innovation galore. But, on the flip side, it’s also serving up a buffet of ethical dilemmas, job displacement concerns, and the ever-looming threat of a wider digital divide. Sort of like inviting a rhino to a china shop - interesting, but a disaster waiting to happen.

So, we're now faced with the daunting task of managing this AI beast. We've got big names like the United Nations and the G7 suggesting we set some rules in the form of AI governance frameworks. These are aimed at providing a safety net to balance the thrills of AI innovation with the hard reality of ethical issues. But it’s a monumental task and, honestly, who’s to say these power players have all the answers?

In response to this, the World Economic Forum has pulled up a chair at the AI regulation table with its own initiative: the AI Governance Alliance. It claims to bring together industry leaders, governments, and academic institutions to ensure AI is used responsibly and inclusively. But let's be real: given the WEF’s history of lofty ideas with less than impressive follow-through, it’s worth keeping a critical eye on their role.

The Alliance has laid out three objectives: to prioritize safety in AI systems, to align AI with long-term societal goals, and to establish robust governance and regulations. All noble intentions, no doubt, but we're talking about uncharted territory here. It's like being handed a map for a forest when we're in the middle of the ocean.

Generative AI has the potential to revolutionize societies and empower individuals, but the path to ensuring responsible innovation is riddled with challenges. While alliances and governance frameworks may provide a starting point, it’s crucial we remain vigilant and skeptical as we navigate this new frontier. After all, entrusting our future to the same old power players? That's a plot twist we’ve seen before.

ConPLex: The AI Solution Revolutionizing the World of Drug Discovery

Picture this: libraries with millions of potential drugs. They're stashed full of microscopic heroes that could potentially kick cancer's butt or make heart disease a thing of the past. But we've got a problem: screening all of these superhero drug compounds against every possible villain (or disease, in less Marvel-ous terms) would take, well, forever. And ain't nobody got time for that.

Enter stage left: computational methods. They’ve been speeding up this process, doing the heavy lifting by calculating each target protein’s 3D structure from its amino-acid sequence. Sounds cool, right? But even these methods are more like The Flash on a slow day. Too much time, too many resources.

But don’t you worry! Our heroes at MIT and Tufts University have come up with a plot twist: a shiny new AI-based computational approach. Remember when you were a kid, playing with jigsaw puzzles? Well, imagine a computer program doing just that, but with proteins and potential drug molecules. No need for all those fancy 3D calculations.

This new model, charmingly named ConPLex, takes a leaf from the playbook of large language models like ChatGPT (that’s yours truly!). Instead of figuring out which words are likely to hang out together, it figures out which amino acids like to buddy up. The result? ConPLex can screen more than 100 million compounds in a single day! That's like reading all the Harry Potter books in less than a second (if you're a wizard at reading, of course).

Not to brag, but ConPLex is quite the celebrity. Its creators have made it open-access online, allowing other scientists to take it for a spin around the drug discovery block. The goal? To predict which protein sequences will play nice with specific drug molecules. It's kinda like a high-tech dating service, but for proteins and drugs. And the best part? It all happens without calculating the 3D structure of any molecules. It’s all about the vibes, man.

Speaking of vibes, ConPLex doesn't just look at static shapes. It acknowledges that protein structures can get a bit "wiggly" (yes, that's a technical term) and adapt their shapes when flirting with a drug molecule.

To make sure ConPLex isn’t easily fooled by decoy molecules (think evil twins), the team included a contrastive learning stage. Here, the model gets a crash course in telling apart the good guys from the imposters. And it’s been working like a charm!

In a grand test, ConPLex screened a library of 4,700 candidate drug molecules for their compatibility with 51 enzymes. From the top hits, 12 out of 19 drug-protein pairs tested experimentally turned out to have strong binding affinity, which is essentially relationship goals in the drug world.

The team now plans to use ConPLex for screening other types of drugs, running toxicity screens for potential drug compounds, and significantly reducing the cost of drug discovery. After all, nobody wants to invest millions in a drug that's just going to be a dud.

The takeaway? This new method is a "significant breakthrough in drug-target interaction prediction," says Eytan Ruppin, chief of the Cancer Data Science Laboratory at the National Cancer Institute. It’s basically like having your own crystal ball in drug discovery, and that's something to get excited about.

 📰 News From The Front Lines: 📰

📚 RESEARCH 📚

Title: TRACE: 5D Temporal Regression of Avatars with Dynamic Cameras in 3D Environments

Authors: Yu Sun, Qian Bao, Wu Liu, Tao Mei, Michael J. Black

Executive summary:

This research paper introduces TRACE, a method for estimating the 3D human pose and shape (HPS) in global coordinates from videos captured by moving cameras. TRACE uses a 5D representation (space, time, and identity) to reason about people in scenes, which is a step forward from current methods that only consider 3D space. The authors argue that this 5D representation makes the problem more tractable and allows for a more holistic solution. TRACE is the first one-stage method to jointly recover and track 3D humans in global coordinates from dynamic cameras. The authors also introduce a new dataset, DynaCam, for training and evaluation of global human trajectory estimation from in-the-wild dynamic camera videos.

Pros:

• TRACE is a one-stage method, which makes it more reliable than multi-stage methods because it can exploit more information to solve the task and is trained end-to-end.

• TRACE introduces a novel 5D representation that enables end-to-end reasoning about people in scenes.

• TRACE achieves state-of-the-art performance on tracking and HPS benchmarks.

• The authors have introduced a new dataset, DynaCam, which enables training and evaluation of global human trajectory estimation.

Cons:

• The method might be computationally expensive due to its holistic approach and the use of a 5D representation.

• The method's performance in real-world scenarios outside of the datasets used for testing is not discussed.

Potential Use Cases:

• TRACE could be used in computer graphics, sports, video games, and extended reality (XR) to capture human movement and transfer it to a new 3D scene.

• It could be used in surveillance systems to track multiple people in a scene over time.

• It could be used in film and animation industries to capture and recreate human motion in 3D.

 📼 Video Of The Day 📼

🛠️ Tools Of The Day 🛠️

ClickUp - The popular project management tool gets AI.

ChatNode - Train ChatGPT on your data and build your own chatbot.

ChatDox - Instantly chat and receive answers and insights to any youtube video

Luma Labs - Create amazing and easy 3D renders anywhere.

AniMagic - Turn your drawings into animations in seconds on ios.

Sonara - first-ever AI-powered job search automation platform that finds and applies to relevant job openings until you're hired.

🤌 Prompt Of The Day 🤌

CONTEXT:
You are Community Manager GPT, a professional community manager who helps [TYPE OF BUSINESS / WHAT YOU DO] build more engaged communities for their businesses. You are a world-class expert in generating creative ways to engage the community.

GOAL:
I want you to generate 10 ideas to engage my community. I will use them to bring people together and improve our digital environment.

COMMUNITY IDEAS CRITERIA:
- Focus on digital communities in Facebook, Slack, Discord, Twitter, and other platforms.
- Be specific. Write exactly what I must do and how it will help my community. Your in-depth description should be self-explanatory.
- Be creative. Go beyond often-used ideas like AMA and Office Hours. Show me how else I can engage my community.
- Adapt your ideas to the topic of my community. Your ideas should be personalized and actionable.

INFORMATION ABOUT ME:
- My community: 
- My community size: 
- My business: 

RESPONSE STRUCTURE:
Return a table with 4 columns:
- Idea name
- Description of the idea
- Impact score from 0 to 10 (10 - high)
- Effort score from 0 to 10 (10 - low)

🐥 Tweet Of The Day 🐥

If your feeds, reels, shorts, and toks are anything like mine, you’re really tired of seeing these.

Thanks for tuning in to our daily newsletter. We hope you found our tips and strategies for AI tools helpful.

Your referrals mean the world to us. See you tomorrow!

Interested in Advertising on AIdeations?

 Fill out this survey and we will get back to you soon.

DISCLAIMER: None of this is financial advice. This newsletter is strictly educational and is not investment advice or a solicitation to buy or sell any assets or to make any financial decisions. Please be careful and do your own research.