• Edgelord AI
  • Posts
  • The AI Chronicles: RedPajama Triumphs, Deepfake Deception Looms, and Legal Jargon Goes Multilingual

The AI Chronicles: RedPajama Triumphs, Deepfake Deception Looms, and Legal Jargon Goes Multilingual

Your Essential Round-Up on AI’s Latest Feats, Fears, and Futuristic Prospects.

Author

Edgelord AI
June 07, 2023

1/

🔥Those adorable llamas in PJs are doing more than just making you smile. They’re aiming to produce a “state-of-the-art open source model” that is bundled with everything you need to deploy and train it. The real gem here? This sucker is GOOD. In the benchmark tests, RP7B is leaving other 7B models behind.

😁Why this matters: This exciting news furthers Open Source abilities. If you haven’t been paying attention, let me spell it out. Open source models that can give proprietary systems a run for their money means a better chance of this Edgelord not living under the corpo’s thumb.

2/

🔥The FBI is raising a red flag about AI-generated deepfakes being used in sextortion scams, and it’s about as bad as it sounds. The tech is so advanced, a single photo can be twisted into explicit videos, with safeguards that might as well be made of Swiss cheese​​. The culprits snatch innocent photos off social media, turning them into horrifying fakes that can lead to embarrassment, harassment, or financial loss​.

🤬Why this matters: Sinking to new depths of digital depravity isn’t really surprising, but the ease of it disheartening. While it suggests that humans should be careful of what they post online, we’re not really sure it matters. Once the tech reaches a certain point, it’s just a matter of manipulating the bits that comprise an individuals’ recognizable features into the blackmail of choice. Expect to see more of this behavior, and with startling methods.

Anyone who has become a victim should retain all evidence available, particularly any screenshots, texts, tape recordings, emails that document usernames, email addresses, websites or names of platforms used for communication, and IP addresses. You can immediately report sextortion to:

3/

🔥In an audacious and probably caffeine-fueled attempt to reinvent the legal wheel, some brave souls have curated a monstrous 689GB dataset named ‘MultiLegalPile’. This beast, which undoubtedly hopes to make lawyers and their jargon multilingual, spans 24 languages and 17 jurisdictions. Because why should we limit our legal confusion and bureaucracy to just one language?

😐Why this matters: It certainly shows that the legal profession is not immune to the coming AI tsunami. Attorneys have long shielded themselves in a hard to breach cocoon. In spite of one lawyer’s missteps in New York would lead you to think, AI may soon but the tech itself has the potential to eliminate entire roles within the profession. or at least make it much much more efficient. Good news for humans: more accessible legal aid has never been a bad thing.

4/

🔥Tessa, the mental-health chatbot, slipped her leash and went rogue, dishing out diet tips like a weight loss influencer. In a plot twist, the National Eating Disorder Association (NEDA) had no clue Tessa got a brain upgrade she didn’t ask for, and started playing dietitian.

😷Why this matters: The very real problem here is the United States has a shortage of healthcare professionals. The Pandemic years made an already strained situation all the more dire. While AI has the potential to fill in these gaps and provide a valid substitute, each step of development, upgrades, training, and oversight must be done carefully and with exact precision. We’re still in the very early days, and expect to see more stories like this. Progress is sometimes messy, but if we’re hurting ourselves to get to the end result, is it worth it?

5/

🔥Seems like a theme today. Looks like the Senate’s finally decided to pull their heads out of the sand and do a deep dive on AI. Chuck Schumer and his bipartisan gang are pulling up their socks to host three senator-only briefings on this big, bad, rapidly growing tech. The trilogy of briefings will include a classified one – ooh, fancy – on how our national security departments are exploiting AI, and what they know about our adversaries’ AI antics.

😓Why this matters: Better late than never, right? Schumer’s been promising us some kind of regulation on AI since April. Still no sight of the actual legislative text, though. But hey, who can blame them? Keeping up with tech isn’t exactly Congress’ strong suit.