201–210 of 620 posts
Forget about 2025. I predict that, despite being a highly valued member of my organization and despite my company making money hand over fist, I will not get a holiday bonus this year.
"Best year ever!" "Sorry, we can't do raises this year." Who is getting all the money?
Earlier quoted context omitted.
> 3) Trade jobs will start to become what code jobs were in the 00's -- very well paid. What signals are there for this?
https://www.mckinsey.com/capabilities/people-and-organizatio...
There are very different business constraints for hiring software developers vs all skilled tradespeople.
Earlier quoted context omitted.
I think Open AI will release a GPT phone and the UX is built around your phone being your personal AI assistant. When you pick it up it's like you are always on a Facetime call with your AI personal assistant. You can skin your AI personal assistant how they look .. to look like a celebrity to a deceased friend or relative whose always there to help & guide you through your day (get things done for you, your knowlegd…
I’m also bullish on this. It looks really cool and revolutionary. And doable. But it also looks horrible in how this will force us to lose what remaining control we had over our privacy :( LLMs are really cool but we really need to make them work locally.
Apple waiting to create such a personal device as noted above for it to work locally will be the decline and possible death of the boring iPhone!
1) Companies will develop many specialized LLMs linked via a router-like app that determines the best LLM to perform your request. This will yield better results without requiring an AGI. It might even be good enough to replace a specialized information tech job. 2) The cost of a line of code will continue to drop. A Moore's like law is coming/here for code. 3) Trade jobs will start to become what code jobs were in t…
Title: "RouteLLM: Learning to Route LLMs with Preference Data"
Abstract: Large language models (LLMs) exhibit impressive capabilities across a wide range of tasks, yet the choice of which model to use often involves a trade-off between performance and cost. More powerful models, though effective, come with higher expenses, while less capable models are more cost-effective. To address this dilemma, we propose several efficient router models that dynamically select between a stronger and a weaker LLM during inference, aiming to optimize the balance between cost and response quality. We develop a training framework for these routers leveraging human preference data and data augmentation techniques to enhance performance. Our evaluation on widely-recognized benchmarks shows that our approach significantly reduces costs—by over 2 times in certain cases—without compromising the quality of responses. Interestingly, our router models also demonstrate significant transfer learning capabilities, maintaining their performance even when the strong and weak models are changed at test time. This highlights the potential of these routers to provide a cost-effective yet high-performance solution for deploying LLMs.
Predictions from 2008: National Intelligence Council/Director of National Intelligence - Global Trends 2025 https://www.dni.gov/files/documents/Newsroom/Reports%20and%2...
> We believe the most likely occurrence by 2025 is a technological breakthrough that will provide an alternative to oil and natural gas, but implementation will lag because of the necessary infrastructure costs and need for longer replacement time. However, whether the breakthrough occurs within the 2025 time frame or later, the geopolitical implications of a shift away from oil and natural gas will be immense.
I wonder if solar's dirt-cheap cost would be considered a "breakthrough". Interesting breakdown on how major OPEC countries will be affected by such a "breakthrough". Great read!
Edit: This prediction is wild also:
(Page 62, pdf page 83)
> A Non-nuclear Korea?
> We see a unified Korea as likely by 2025—if not as a unitary state, then in some form of North-South confederation. While diplomacy working to end North Korea’s nuclear weapons program continues, the final disposition of the North’s nuclear infrastructure and capabilities at the time of reunification remain uncertain. A new, reunified Korea struggling with the large financial burden of reconstruction will, however, be more likely to find international acceptance and economic assistance by ensuring the denuclearization of the Peninsula, perhaps in a manner similar to what occurred in Ukraine post-1991. A loosely confederated Korea might complicate denuclearization efforts. Other strategic consequences are likely to flow from Korean unification, including prospects for new levels of major power cooperation to manage new and enduring challenges, such as denuclearization, demilitarization, refugee flows, and financing reconstruction.
Edit 2:
(Page 75, pdf page 95)
> Potential Emergence of a Global Pandemic
> Experts consider highly pathogenic avian influenza (HPAI) strains, such as H5N1, to be likely candidates for such a transformation, but other pathogens—such as the SARS coronavirus or other influenza strains—also have this potential.
...
> If a pandemic disease emerges, it probably will first occur in an area marked by high population density and close association between humans and animals, such as many areas of China and Southeast Asia, where human populations live in close proximity to livestock. Unregulated animal husbandry practices could allow a zoonotic disease such as H5N1 to circulate in livestock populations—increasing the opportunity for mutation into a strain with pandemic potential. To propagate effectively, a disease would have to be transmitted to areas of higher population density
> Outside the US, critical infrastructure degradation and economic loss on a global scale would result as approximately a third of the worldwide population became ill and hundreds of millions died.
At least we didn't get hundreds of millions dead. That's pretty grim.
-USA stops supporting Ukraine war efforts and Ukraine is forced to cede large portions of land to Russia, creating a DMZ along vast stretches of the new border. -Expanded land use restrictions and federal tariffs against solar and wind, resulting from pressure from fossil fuel lobbyists and the new head of the department of energy. -Another CEO is shot. -Crypto has a major correction; bitcoin finishes the year well u…
Non-anthropomorphic generic AI models.
Impact of war on prices of all kinds of stuff everywhere will become more evident.
Dopamine management medication will lead to a renaissance of human ingenuity and scientific discoveries and a big crash for social media and gambling sites (and in fact most online activities) but everyone will start becoming more robotic and even less social. Think GLP-1 inhibitors except for thinking more deeply and for longer periods. If not 2025 very soon.
Earlier quoted context omitted.
When all knowledge workers and creatives in the US are out of jobs in 2 years because of AI, those trade jobs aren't going to pay well because no one's going to have any money to pay for any work. No new houses being built, no expensive kitchen remodels, etc.
Do you know just recently there was no internet or computers and there were still high paying jobs for smart people?
Google will announce new products (will call them initiatives) and will not launch them to the general public. ChatGPT Search will get significant (high 1 digit) share of the search market.
I'll bandwagon on chat.com getting 10% or higher marketshare. Look back at Google search's 800 number; https://techcrunch.com/2013/05/12/google-kills-sms-search/ (the only article I could find). We're at that stage of chat.com going mainstream. Already chat.com is a better experience for finding local shops and restaurants. ChatGPT voice mode was surgically amazing while driving. I had a personal assistant refining c…