In a rapidly evolving digital landscape, the early rollout of ChatGPT’s advanced voice mode has captivated tech enthusiasts worldwide. This groundbreaking feature allows users to engage with AI in more realistic and responsive ways, opening up new possibilities for everyday interactions. But that’s not all; this article will delve into other intriguing stories emerging from the world of artificial intelligence, including predictions about the arrival of superintelligence, a significant update from Google, and a free tool that everyone should explore. Stay tuned, as there’s much to uncover in this tech revolution.
The Rise of Superintelligence: Are We Ready?
Recently, tech visionary Sam Altman released an essay predicting the imminent arrival of superintelligence, spurring debates about its impact on society. He estimates that this could occur within the next 2,000 to 5,000 days, roughly between 2030 and 2038.
Altman discusses the necessity of infrastructure to support AI’s growth, warning that a lack of resources could lead to a situation where superintelligence becomes a limited commodity, potentially sparking conflicts over access. Read more about Altman’s predictions here.
This framing raises crucial questions about the future educational landscape and the role AI will play in our lives. While many anticipate AI’s capacity to solve pressing global issues, Altman emphasizes the need for robust infrastructure to avoid unequal access. As superintelligence approaches, the balance of knowledge and power may shift dramatically.
OpenAI’s Ambitious Data Center Plans
In a related story, OpenAI has revealed plans to build five to seven massive data centers that could each require as much energy as five nuclear reactors. Collectively, this could provide enough power for millions of homes, inciting skepticism about the feasibility of such grand ambitions. Experts express concern that even with significant computing power, the journey toward superintelligence is fraught with challenges. Learn more about OpenAI’s data center plans here.
The urgency for infrastructure grew from the realization that AI’s potential may be hindered by its energy demands. In light of this, Microsoft has formed strategic partnerships to explore energy solutions that would facilitate the necessary growth towards superintelligence. Discussions around societal implications continue as the world observes these ambitious developments.
Google’s Gemini 1.5 Pro Update
Meanwhile, Google debuted the Gemini 1.5 Pro, an upgraded AI model designed to boost performance while reducing costs. It boasts an impressive capability to analyze up to 2 million tokens, making it a valuable tool for complex data handling and mathematical tasks. However, the peculiar naming conventions of these models, like Gemini 1.5 Pro2, leave many wondering what future iterations might entail. Find more details about Google’s Gemini update here.
While Gemini 1.5 Pro shows promising improvements in benchmark performance, it still trails behind OpenAI’s latest iteration. The comparison highlights the race between major players in AI to establish their dominance in developing advanced models that can handle real-world complexity with sophistication.
Google’s Notebook LM: A Game Changer for Users
In another significant announcement, Google introduced Notebook LM, an innovative free tool that allows users to interact with their documents in conversational formats. Users can upload PDFs or text files to generate engaging readings or discussions about the content. This tool promises to revolutionize how individuals and professionals approach document engagement. Explore Google’s Notebook LM here.
As AI continues to grow and adapt, this tool could democratize access to information and streamline workflows, particularly for those who find traditional document review tedious. Its simplicity and effectiveness have garnered praise, suggesting a step forward in user-friendly AI applications.
Final Thoughts
The breakthroughs highlighted in this article underscore the promising yet complicated journey of artificial intelligence and its growing integration into everyday life. From the potential of superintelligence to new tools like Google’s Notebook LM, both excitement and skepticism abound. Which of these advancements do you find most significant for the future? Share your thoughts and engage in the ongoing conversation about AI’s role in our world.
Discover more from Marki Mugan
Subscribe to get the latest posts sent to your email.
Why did I think the table was huge, or that Jerry was really very small? Am I hallucinating?
shame google will shutdown notebookLM before the end of the month
Yes , I was sufficiently interested that I actually subscribed.
Chat does do link anymore…not happy
Imagine trying to create an accurate world model in your head if your only input of the outside world is via text. its like trying to reverse engineer reality based on only how people describe the world using the abstraction of symbolic language. an uphill battle, but these models do well, its a testament to how descriptive certain human languages are. these models need to be trained from the ground up with multimodal inputs, than we will truly have no chance…
I'm using notebookLm since it's beta stage amazing tool
Notebooklm is impressive. I have started generating "podcasts" of subjects that I want to get an overview over.
I tested NotebookLM with a novel I'm working on, about 340 pages. It got quite a few details wrong. It's still a work in progress.
7:28 where he talked about that newly released paper ?
Notebook LM took my by surprise.. is awesome.
The fact that it doesn't have any real time data is disheartening. I believe it did have it in prior versions but for some reason they took it out. Maybe for privacy reasons?
I didn’t imagine a tilted table. I imagine a vertically thick table whose top surface was at shoulder height and whose bottom surface was ankle height.
Strange that OpenAi chose to name its voices after the new range of Scarlett Johanssen's bathroom air fresheners – Breeze, Cove, Ember, Juniper, Arbor..🤦♂️
From a quality or clarity perspective I'd say Elevenlabs is on par with these GPT 5 voices.
It seems like the wording of the question should be that he puts the cup upside-down against the table, instead of on, but maybe that makes it too obvious.
9144 Dell Lights
personally i suspect that as the AI becomes intelligent, which is different from knowledgeable, it will only keep the intelligence it found in the big piles of Human-sourced documents and archive those documents or just keep a pointer to those documents. and my second intuition is, you could train a super AI from just the works of 1 or a few humans. So maybe the super-AI will in-fact necessitate LESS power than our chatBots bloated with tons of unimportant things they could google and give us better answer than re-creating it using the 'what token should a output next ' technique. but but if there's 1 billion users, some humans, some robots … then well that's your new nuclear reactors
I felt the same as you about Notebook LM – I was very excited about it, loved the concept — the audio was quite something. I had a real world use case, but once I tried to use it, it very quickly went down hill. It soon became apparent that the tool was a demo or "proof of concept" rather than actually usable. I was really disappointed, in the end! Here's what I found:
– You can't collaborate with or share a Notebook if you set it up using a Google Workspace account.
– If you add a source from a URL or PDF the text is extracted and the URL/PDF is lost. It is therefore impossible to follow a citation back to it's source web-site or doc. This makes it unsuitable for research.
– Maximum of 10 citations per summary restricts what can be included in a summary.
– If a summary is created, and a change is requested (e.g. add new items to a list). The new summary created has citations for the new items, but all previous citations are lost.
– No way to create a list of all sources, or even a list of the sources cited in a summary
– No way to export, publish, share, or get any information back out of NotebookLM – e.g if I want to share the summary, email it to someone – publish to a web-site, post it on social networks, you can't
– Chat will often provide responses that objectively wrong- e.g. ignoring sources provided. I have to tell it which source to look at, and what the source says. Only then will it use it.
– Chat history does not persist
– Gives nonsensical answers to questions. I asked it about something specific relating to UK – it gave an answer about a topic in Kuwait
– Ignores specific aspects of my questions, summaries are oversimplified and "dumbed down". Supposedly runs on Gemini 1.5 Pro, yet I get far better answers/nuance from Gemini 1.5 Pro in AI studio if I use it directly with the same source material.
– I can't take over and edit the summaries, they're read only.
Can you do a Video comparing strongest Chinese A.I models and robotics Compared to Western Counterparts ?
The person could be laying on a level table.
I want a voice chat AI naruto and i want him as a therapist LOL
One consideration is that compute to train the model is much different from compute to run inference. Inference takes orders of magnitudes less compute, so it's likely the training costs will be aggregated into the costs of providing the answers ( not also forgetting the costs of researching ways to improve the model in ways beyond just adding more data. )
This is a stretch to ask, you prolly have it in your paid section – but just found out myself about Liquid AI and their form of LLMs. Hopefully you can touch on those in the future.
This money's for your latest vid because it's not enabled on that one. I thought I followed the news you covered but you caught a LOT I missed.
Notebook LM can do a lot more than just the podcast function. It is brilliant at summarizing documents or formatting tasks, no hallucinations, spends as long as needed to fully execute the prompt! Google is finally onto something!
Sam AltMan is funny but early is a playful of the day in my opinion 😅
so Americans will be deporting allegedly millions of people and at the same time some people would have to remember to live in the context of Modern life rather than 1798 and then who can build nuclear reactor unless they have 40 years of time to waste? 😮😮😮😮
regarding the strawberry reasoning task you created, couldn't the guy have had his ankle lifted in the air so as to touch the bottom right corner with a flat table?
13:17 doesn’t that test comparison between the two models only suggest that one had more training on Matthew Bermann’s videos than the other model?