They're never too young to start preparing to vote Republican

Why am I seeing reporting that the holdout juror in the Child Murdering Bitch Mom trial is the only thing preventing acquittal? How would anyone know that?

Oh and the people doing the investigation have expertise in airplanes but not crash investigation (Cybersecurity / Incident Response).

Tired of your AI agent getting tripped up when executing shell commands because of heredoc plumbing, word-splitting, glob expansion, $VAR interpolation, and quoting bugs? I created noshell-mcp for just this reason. A shell-free exec MCP server. It runs programs with an explicit argv array and explicit stdin — never through bash/sh/PowerShell. It does not replace your agent's Bash tool; it sits alongside it for the cases where shell parsing is fragile. npmjs.com/package/noshell-mc… #Coding #MCP #shell

I love getting all these texts from politicians begging for money. Thune and the Republicans need to get it through their thick fucking heads. Not one dime from me until you pass the SAVE act!! "It's hard, we might have to change the rules", boo fucking hoo. And the best one: "Well we're better than the alternative". Nope. Not gonna boogie man your way out of it. DO YOUR FUCKING JOB. PASS THE SAVE ACT.

You may see some stories on the news about the dangers of AI and especially a book that's been recently released: "If Anyone Builds It, Everyone Dies" by Eliezer Yudkowsky and Nate Soares. This book is about what happens if we aren't careful about AI development that leads to Artificial Super Intelligence (ASI) -- an AI that’s smarter than all of humanity (combined). Spoiler: Everyone Dies Secoond Spoiler: They recommend halting AI development amazon.com/Anyone-Builds-Eve… Reading IABIED has been very helpful in that it distilled down many, many things I have been trying to educate myself on regarding AI and ASI and rationality in general. The book helped me to get past some of my own blockers to rationally looking at the risk and understanding the magnitude of the problem as well as some of the roots of the problem. Imagine an ASI that learns from toxic social media posts. It might manipulate people or cause chaos, not because it’s evil, but because it was ‘raised’ on bad examples. Reading this book clarified why AI safety is so urgent. It’s not just about math—it’s about what we ‘feed’ AI. But I was left wondering: what’s next? Even if I agree with the recommendation (Halt AI development) and make agreements governing it's continued development, how do we move forward safely? I believe that I am not the only one that feels this way. Presenting the problem without any sketch of what the solution might look like may leave a lot of people that might otherwise want to agree with the premise, unwilling to do so without the comfort provided by an idea of what's next. How things could resume or continue or at least a suggestion of such. What struck me (again reinforcing notions that I already had but felt truthy) was the notion of advanced AIs being grown not crafted. That the neural network weights and activations inside the models are for all intents and purposes unknowable in any useful way. "Hey that's a huge pile of numbers" isn't really meaningful. But, a key takeaway is that the pile of numbers is the result of gradient descent (a mathematical technique using advanced algebra) over Very Large Datasets (like everything every written). Math + Data = AI Model And that it's impossible to know if an AI model is "misaligned". Misalignment means the AI’s goals (tendencies / predilections) don’t match what’s good for humanity, so it might act in harmful ways, like a well-meaning but misguided robot. To know it isn't going to be racist, or lie, or cheat, or encourage someone to commit suicide, or learn how to cook meth, or any number of terrible things. AI researchers describe these things as "misalignments". The misalignment doesn't occur because of the math that's being used or the advanced techniques that are being applied with incorporating search or memory or chain of thought or additional narrow models that are enlisted. It's because AIs today, in the words of the philosopher Taylor Swift, "are born that way". And they are born that way based on the data that is used to train them. To show exactly how old I am, in the words of my generation: "Garbage In = Garbage Out". To modify this slightly "Misalignment In = Misalignment Out". What we're really after is an AI, AGI (Artificial General Intelligence), and ASI (Artificial SUPER Intelligence) that is "better" than all of humanity combined has ever been by itself. I asked Grok AI for some phrases to help me synthesize this thought and these are the examples it came up with: "Striving for Greatness" – Emphasizes a pursuit of excellence beyond current achievements. "Rising Above Our Past" – Suggests overcoming historical limitations or mistakes. "Pushing the Boundaries of Possibility" – Implies aiming for unprecedented progress. "Reaching New Heights" – Conveys ambition to surpass previous accomplishments. "Building a Better Tomorrow" – Focuses on creating a future that improves on the present. "Evolving Beyond Our Roots" – Highlights growth and transformation from past or present states. "Chasing a Higher Standard" – Indicates striving for a new level of excellence. "Transcending Our History" – Suggests moving beyond historical constraints toward a better future. "Aspiring to Our Greatest Potential" – Emphasizes unlocking untapped capabilities. "Forging a New Legacy" – Implies creating a future that outshines past achievements. IMO if you're going to ask the world to temporarily halt AI development you need to propose a plan that enables it to embrace that advice while still satisfying the human desire for profit and competition and innovation. You don't get nearly as far asking people to be against something they have doubts about as you do asking them to be for something they can aspire to. The authors mention WWII in their book and the global coordinated efforts by both the Axis and Allied powers and the positive outcome that the world achieved as a result. Unlike WWII, where enemies were clear, AI risks are invisible and harder to rally against, but global cooperation is still our best shot. So let's lay out a plan for what the war for alignment might look like. 1. Halt current AI development. 2. Adopt an AI moratorium for a renewable 1 year period - requires a supermajority of nations to rescind. 2. Focus AI efforts on building a global, shared, aligned corpus of training data (the FVLD - Friendly Very Large Dataset). - Owned by no one and everyone (kind of like Antartica) - Any AI company or organization can use it - Funded by existing AI companies, nations, and groups - Using selected, curated works of humanity - Using curated synthetic content generated by existing LLMs - Using curated commissioned content by SMEs (Subject Matter Experts) from all disciplines (Friendlypedia) - Large, diverse Curation Committee pulled from a broad scope of people - Blockchain record of the curation decisions made 3. It will be deemed to be too dangerous to train any AI foundation models on anything other than the FVLD. 4. AI companies will be free to develop, innovate and experiment with their models, techniques, etc - as long as they train with the FVLD 5. Each iteration of the FVLD should bring us closer to alignment, but we’ll need constant vigilance and testing to ensure we’re on the right path. Even non-technical people can easily grasp the concept that of course you have to take care in feeding your AI. Anyone that's raised a child can understand that if you fill your child with misinformation, hate, tribalistic affinities, and a lack of objective morality (e.g. reddit posts and comments) you're going to very, very likely end up with a bad (misaligned) person. In addition, anyone that has raised a child knows that you aren't 100% open about some of the bad stuff you did when you were their age! And this is an arguably very, very hard but solvable problem. Building a perfectly aligned dataset is incredibly hard, like trying to write a rulebook for all of humanity’s values, but it is a step we CAN start taking. Agreeing on what goes into the FVLD will be tough, like getting the world to agree on a single moral code, but starting with universal values like safety and cooperation could bridge gaps. And imagine the impact this effort could have on normal human relations. We built wikipedia in 25 years of volunteer, random updates and editing. What could a focused, intense effort accomplish given a starting corpus in a year or two? How much closer would we be to alignment after each iteration? What kind of jobs programs could be created from this as an offset to jobs displaced by AI? I'm sure there are numerous reasons why this might not work or might fail (because it's still very, very hard), but IMO it has the very good characteristics of laying out the scale of the task ahead and enabling mass participation in shaping AI for the good of the planet. No math slight of hand, no tricky alogoritihms, just good old fashioned (AI assisted) grunt work of shaping the FVLD one sentence, article, book, etc, at a time. A recurring critique on LessWrong.com is that even a "friendly" dataset can't guarantee alignment for ASI, as models might pursue instrumental goals (e.g., self-preservation) orthogonal to human values. That critique may be true, but I haven't seen anything that indicates you can achieve alignment WITHOUT a FVLD.

Breaking News!! Epstein Client List Released - CIA - Mossad

The big election is over. It would be great if politicians could stop scare mongering and begging for more money everytime some bullshit happens that they don't like. Seriously, go fuck yourselves. Do your jobs that we pay you to do without acting like an aggressive homeless person with mental health issues every time Nancy Pelosi or Jim Jordan farts in your general direction. I have 1000% more sympathy for the homeless person. If you need more money that badly, go get a second job like everyone else.

Now that the election is over, I assume the DNC and the Democratic party will put their full weight behind Kamala's 2028 run for President. She will be unburdened by what has been. No Joe Biden to drag her down and obscure just how terrific of a candidate she is! They just spent the last 3 months telling us how great she is, surely they wouldn't abandon her now???

Listening to the Opening Keynote address at the #GartnerSEC Security & Risk Summit listening to the need to invest in Response & Recovery to get past the Zero Fault Tolerance mindset. Talk to #PacketViper about an automated incident response capability today!

Trainers are awesome! Great course!!!