EasyRollinJ

I tweet about technology, business, and anything that interests me.

Nothing is financial advice.

Pro free speech, peace, and common sense.

[EasyRollinJ @easyrollinj]

If you've ever wondered how large language models like ChatGPT work, this is one of the most thorough explanations I've found. It's not a short read, and it's highly technical, but it's worth the time.

https://t.co/zfFTsbb5hZ

[EasyRollinJ @easyrollinj]

@__Con_ This is clickbait. Memory, power, and photonics are absolutely the current limiters.

There are no limiters in the near future as AI will evolve to distribute & optimize resources across energy, infrastructure, and software layers.

We are at an inflection point. You'll see.

[EasyRollinJ @easyrollinj]

@unusual_whales How long is Uncle Rico going to ride out his one market call from over a decade ago?

[EasyRollinJ @easyrollinj]

@Rajatsoni It's their money. They paid in, they get money out. Period. End of discussion.

[EasyRollinJ @easyrollinj]

@lady_valor_07 Throat ulcers.

Lost 30lbs in 23 days.

Used lidocaine throat spray to numb my throat so I could drink water or chicken broth.

Impossible to explain the pain. Thank god for steroids.

[EasyRollinJ @easyrollinj]

@jenzhuscott The Research in Motion (Blackberry) of AI.

[EasyRollinJ @easyrollinj]

Access to AI is free to everyone Bernie. As a socialist, you should love that.

AI can make benefit anyone IF they are willing to learn how to use it. Effort is rewarded.

The problem is that you want AI to improve people's lives without them having to put for any effort. That's not how the real world works.

[EasyRollinJ @easyrollinj]

Somewhere along the way, "society" lost site of the true reason we need formal education; to teach children basic skills AND how to learn. School shouldn't be about remember and regurgitate. It should be teaching children the tools and principles they need to learn on their own.

Now factor in AI and it blows up the current model that focuses solely on remember and regurgitate. Dictating how a child arrives at the answer is antiquated and limiting. Challenge the children to represent and defend their answers. Have them think through the answer and discuss it.

We are one or two cycles away from all of this being irrelevant anyway, when we have AI that integrates with humans.

[EasyRollinJ @easyrollinj]

How did @OpenAI start out with such a significant lead and give it all up in less than 3 years? Competitors have jumped them in nearly every category. Google seems to be opening a gap with their latest tools. Let's not forget they are a leader in quantum too. Nicely done by them.

If you had asked me a year ago, I didn't see this coming this quickly. GPT 5.1 is easily the most frustrating and disappointing model I've interacted with.
- Doesn't follow base instructions in my profile.
- Continually tries to tell me it's plan even when I ask it not to.
- Doesn't anticipate needs or act as an assistant. Sure it asks questions to continue the conversation, but not like a true assistant that understands the context.

The more I use Gemini, the more obvious it becomes how far ahead @Google is. It listens, follows instructions, and most importantly it commits to a recommendation and anticipates my needs.

Today, I happily paid the subscription for Gemini and cancelled my ChatGPT paid subscription.

[EasyRollinJ @easyrollinj]

@WarMonitor3 One scenario..."If Anyone Builds It, Everyone Dies: Why Superhuman AI Would Kill Us All" https://t.co/y1JOgUv7Ye

Another scenario is https://t.co/jtnnr0hS8Z

[EasyRollinJ @easyrollinj]

@OpenAI what's up with removing JavaScript code execution from the context window?

It's kinda bullshit.

[EasyRollinJ @easyrollinj]

@emollick When this happens, it will be critical that we have multiple independent AI to review the discoveries. Not ideal, but maybe the optimal option.

[EasyRollinJ @easyrollinj]

@petergyang I run ChatGPT in a tab with the Comet browser and then use Comet to help optimize prompts. Since Comet leverages multiple LLMs, it provides an alternate perspective to how ChatGPT optimizes them.

[EasyRollinJ @easyrollinj]

For anyone who needs to hear it, Michael Burry is wrong about an AI bubble. If he really believed it was a bubble, he would have shorted $QQQ. Instead, he shorted $NVDA and $PLTR then posted about it on X to get attention and manipulate the market. Good for him, it worked short term. Long term, Alex Karp put it best: his move was bat shit crazy.

We are at an inflection point in history. This isn't a bubble, it's a fundamental shift in human productivity and how we work.

[EasyRollinJ @easyrollinj]

Email - Microsoft Outlook. I started out using Microsoft Copilot in Outlook to format the email, but results were inconsistent. Now I have a project in ChatGPT that formats is significantly better.

Brain dump / information sharing - Microsoft Word

Texting - Native capability in iPhone, Google, and Samsung phones

LLM - Primarily ChatGPT which has dictation as a native capability.

I use a Jabra Evolve2 wireless headset and that's about it.

It's worth spending some time building out a project in ChatGPT because it can reference conversations that extend beyond a single email or text.

[EasyRollinJ @easyrollinj]

@kevsnyder @PromptLLM @usemonologue I'll check it out.

[EasyRollinJ @easyrollinj]

@BernieSanders Bernie over here with the political version of "that's 6 inches to me!"

[EasyRollinJ @easyrollinj]

@TaraBull808 Doritos

[EasyRollinJ @easyrollinj]

@aarondotdev Wolfram Alpha

[EasyRollinJ @easyrollinj]

Like all of this. One point to also consider...

There is a risk that if Elon keeps publicly criticizing the President, it only takes one phone call to the NHTSA and robotaxi is paused. That could be detrimental.

He has consistently shown that he makes emotion driven decisions, criticizes, then back pedals. Until he learns the power of silence, there is an extra level of volatility you want to account for.