gemini

对 #GPT-5，Google的最佳策略可能是推出 #Gemini-2.6，轻轻压过。

#Gemini 2.5有一个奇怪的bug，他始终认为最新的Gemini是1.5。

#Gemini 何时反击？

#Gemini #Think (2.5 & 2.0)经常想着想着就忘记了System Instructions。

免费用户的 #Grok-3使用限制缩紧了很多，几句话就用完了。好在我现在有 #Gemini 2.5，还是免费无限量。Google大气！

AI疯狂发展的现在， #Gemini 2发布至今已经霸榜一年了，简直匪夷所思。
https://lmarena.ai/leaderboard

还是 #Gemini 的 #DeepResearch 最靠谱，即使只有免费的2.5 flash。

Judge Mehta's decision in the Google antitrust case is so bad I think he should be prevented from working in the law in any capacity. He just threw off his clear responsibilities and let an illegal monopoly maintain its illegal monopoly and continue harming everyone, based on "reasoning" that a child could debunk. It's embarrassing, it's cowardly, it's self-contradictory, it's lawless, it violates precedent, and it will harm countless people and businesses in the tech sector. A staggering achievement really. It even violates Supreme Court precedent, which dictates that judges must apply remedies that end illegal monopolies (and, I believe, confiscate the illegal gains, though I am less clear on whether that's Supreme Court precedent).

Here's an except from a post on Matt Stoller's BIG newsletter, which is very good on the subject of antitrust:

The last meaningful reference point for an antitrust remedy is the Microsoft case. In that one, the break-up was overturned, and a weak interoperability mandate was imposed. But the real penalty to Microsoft was embarrassment and fear within the executive suite; no longer would the company crush its rivals, from then on, lawyers would cautiously oversee product design. That’s not ideal, Microsoft should have just been broken up and set free to compete. But a chastened leadership did have the effect of not killing the next generation of companies, who ended up creating Web 2.0. That’s deterrence, which is one goal of antitrust remedies.

This remedy, by contrast, is obviously going to fail. And the main reason is that, unlike Microsoft, Google’s leadership is utterly unchastened. Google CEO Sundar Pichai and chief legal officer Kent Walker will get bonuses for what they did. They see this conflict as one in which they fought bitterly, and kept at it, and shredded documents, and the result was… victory. They will have no compunction continuing to engage in unlawful behavior. After all, what’s the worst that could happen? Would a rival or the government really go before a weak judge who doesn’t want conflict, and convince him to act? I don’t think so. In other words, this decision isn’t just bad, it’s virtually a statement that crime pays.

(emphasis mine)

Stoller recently wrote a post titled "Why Is Google Still in One Piece? The Terminating a Monopoly Problem" with the subtitle: "Google has lost three separate antitrust cases, and more are on the way. Why does this company still exist in one piece? It shouldn't, but we're still dealing with the hangover of the 1990s."

The problems with the tech sector go all the way to the tippy top.

#Google #monopoly #IllegalMonopoly #antitrust #dev #tech #Gemini #android #Chrome

用 #Gemini 写了个 #国际象棋 教学程序，不过水平很菜

#Gemini 1.5 Pro 002确实比其他的Gemini都慢得多，但不总会给出最好的输出。大概类似 #o1 ，用了更多算力来推理，结果是倾向于较短的输出。

#Gemini #DeepResearch 可以研究用户提供的私有内容了。

🚨 BREAKING #Google just activated #Gemini on #Gmail - without asking you.

Turn it off now; here's how!
https://tuta.com/blog/how-to-disable-gemini-on-android

✊️ Fight AI & fight Google

You have to manually turn off Smart Features in the Setting menu in TWO locations.

Share so everyone is aware. ❤️

AI生成视觉作品，最大的难点在于没有 #反馈，他顶多只能想象而无法看到自己写的代码运行的结果，本质上是在盲画。这个条件下，我相信 #Gemini 3已经超越了所有人类。

Hah! I knew it. I knew the fediverse would pop up as I explore Gemini. Adding this to my collection of platforms to try.

https://github.com/dimkr/tootik/releases/tag/0.15.3

#Gemini #Gopher #HomeLab #ActivityPub #Fediverse

#o1 和 #Gemini Thinking答案一样的话，我基本就不再怀疑了。

#DeepSeek 还不支持多模态，和 #Gemini #GPT 还没法同台竞技。

#GPT-5果然不行，怪不得 #Gemini 连2.6都懒得出了。

有代码辅助的数据分析能力， #Claude 3.5 Sonnet (New)> #GPT-4o >>> #Gemini 1.5 Pro 002。
这大概也是Gemini可以免费无限使用的原因之一吧。
不过免费的Claude和ChatGPT都只能分析两三次就超限了。

#Gemini 果然更新了gemini-2.0-flash-thinking-exp-01-21，支持1M tokens和Code execution，默认Temperatures是0.7。

> 给一个不会方程的小学生讲下面这道题，应该怎么讲？有四个数，第一个数是另三个数之和的1/2，第二个数是另三个数之和的1/3，第三个数是另三个数之和的1/4，第四个数是26。这四个数之和是多少？
这个 #小学数学题 ，难倒了 #GPT-4o、 #Claude 3.5 Sonnet、 #Gemini 1.5 Pro 002。
不说不用方程，基本都能算对。限定不用方程，一个个都傻了，答案千奇百怪。
那些 #数学基准跑分 都是怎么跑出来的？

#Gemini Experimental 1114可能就是 #Gemini2 了，传言11月14号发布不是空穴来风。

感觉 #Claude 已经被 #Gemini 远远甩开了。

I can't explain it, but browsing via the Gemini protocol has a calming effect on me? I'm focused on the content. Found a Gemini front end for Wikipedia and it's clicking for me. No distractions, just content + links.

> I have no idea what I'm doing but it's fun to poke at new things.

#Gemini #NotThatGemini #SmallWeb

Regarding the last couple boosts: among other downsides, LLMs encourage people to take long-term risks for perceived, but not always actual, short-term gains. They bet the long-term value of their education on a chance at short-term grade inflation, or they bet the long-term security and maintainability of their software codebase on a chance at short-term productivity gains. My read is that more and more data is suggesting that these are bad bets for most people.

In that respect they're very much like gambling. The messianic fantasies some ChatGPT users have been experiencing fits this picture as well.

#AI #GenAI #GenerativeAI #LLM #tech #dev #ChatGPT #GPT #Gemini #GamblingAddiction #nihilism

用 #GeminiCLI ，几乎耗尽了三个账户的免费 #Gemini API配额，OCR出来了秦晖的《 #拯救民主 》。虽然还不完美，不过也很接近了。