So many AI and SaaS platforms get it wrong
They really think that millions of people are going to spend hours learning how to create great images, text and video.
Nobody – apart from a few early adopters – is going to do that.
If you want to provide a tool with lots of features, go B2B. Focus on creating an ecosystem of agencies and consultants who can get a return on their investment in a learning curve.…
The real AI thing will be some kind of AI agent that can handle all of your online digital interactions, including but not limited to:
– Managing the agenda,
– Writing and answering emails,
– Web browsing and online transactions,
– Summarising and commenting on news stories
– Taking part in meetings and phone calls, etc.
Imagine having an AI agent that efficiently takes care of all your online interactions, meeting with it for 20-30 minutes a day to give further instructions or get updates, and having the rest of the day to think or do whatever you want.…
While the performance of GPT-4 Turbo and other closed AI models is improving, open-source models are gaining ground quickly. Recently, China’s 01.AI released an open-source model, Yi-34B, that outperformed its open-source peers on several benchmarks. On MMLU, a benchmark measuring basic knowledge across 57 subjects, for example, Yi-34B scored 76.3%, surpassing the 68.9% and 70.4% of other open-source models like Meta Platforms’ LLaMA2-70B and the TII’s Falcon-180B, respectively.…
Three models of Claude 3: Haiku, Sonet, and Opus.
Haiku is noted as the fastest but less accurate, suitable for customer service. Opus is the most powerful, designed for tougher logic questions. Sonet is in between, serving as the free version. Opus outperforms GPT 4 and Gemini 1.0 Ultra in various benchmarks. Sonet even outperforms GPT 4 in some cases. Claude 3 now includes vision capabilities, surpassing some leading models in certain tasks.…
People expect ChatGPT to be as accurate as an encyclopaedia and as comprehensive as Google. But that’s a misconception. ChatGPT is trained to create content, not to tell the truth.
GPT models require fine-tuning, embedding, plugins, etc. on top of the base model to provide reliable information.
The real added value of ChatGPT lies in the language skills: for the first time you interact with a machine in natural language.…
Hopefully, sooner or later, people will realise that custom GPTs are not exactly the way to achieve AGI.
Actually having to tell the model which other model to use to perform certain tasks is the exact opposite of AGI.
Among many other things, AGI is solving all your problems online, learning from mistakes, exploring new creative and efficient ways to deliver value, interacting with you as a conscious independent being, etc… and all of this implies that you do not tell the model how to do things, even at an abstract semantic level.…
Everyone is confused and stressed about what to build on AI because everyone is focused on a moving target: technology.
No one is talking about what companies really need. The real challenge is not the AI models, the blockchain, the apps and so on, but how to create value.
Technology is a moving target and will remain so. Its only part of the equation: the fundamental part is people.…
The explosion of AI startups in dozens of sectors masks something many of them share: They are increasingly built on top of standardized technology from a few AI giants like OpenAI, Google, and Meta. This puts a premium on strategy over proprietary technology.
https://hbr.org/2023/12/strategy-not-technology-is-the-key-to-winning-with-genai
In the interview with Sam Altman, he discusses the unexpected success and transformative impact of ChatGPT and GPT-4, emphasizing their greater usefulness than initially anticipated. He provides insights into GPT-5, highlighting its increased intelligence and capacity to handle longer, more complex problems. Altman envisions a shift towards natural language interfaces for computer interactions, where workflows are conducted inside language models.…
In this interview, Roger Penrose, a mathematical physicist, discusses his interest in consciousness and its departure from his usual scientific pursuits. His fascination emerged from early discussions with his father about the computational nature of consciousness. During his undergraduate studies, exposure to Godel’s theorem and discussions on quantum mechanics further shaped his views.…
It is like SAP inviting people to build and sell their own ERP based on SAP. Makes sense if its focused on custom knowledge for specific business cases.
But people interested in custom GPTs are mostly companies and professionals. And companies need reliable and qualified custom services, not just “plug-ins of expertise”.
It will all depend on implementation, but I wouldn’t try to be all things to all people.…
It is important to note that what most people are using (the gpt-3.5 version) is probably a dumb gpt-4, not an older version of the model. I have no way of proving this, it is just an intuition given the results of this and other tests.
For this test, we want to test both the hallucinatory bias and the ability to calculate of the gpt model. The best way to do this is to ask the model to do some arithmetic according to an abstract rule (see the code above).…
Todos los grandes modelos de inteligencia artificial generativa están siendo censurados, tanto a nivel de empresa (entrenando el modelo para dar sólo respuestas políticamente correctas) como a nivel de estado, donde la regulación europea es sólo el comienzo.
Por un lado habrá modelos centralizadosm, entrenados con “información procesada” para consumo masivo.…
Almas de Metal, a caballo entre 2001 y Blade Runner, antecedente de Terminator. Una obra de arte de la ciencia ficción, recreada como serie hace muy pocos años por HBO con Anthony Hopkins en “Westworld”.
El hombre en su esquizofrenia filosófica afirma no haber sido creado por nadie, y al mismo tiempo cree poder fabricar máquinas conscientes y por lo tanto creativas.…
IMF Report: AGI destroys all jobs within 5 to 20 years! Frontier of Automation expands beyond humans
Anton Korinek
Fellow, Brookings Institute
Professor, UVA
Former, Johns Hopkins, IMF
Frontier of Automation – Task complexity of machines increases over time
Unbounded Distribution – Human task complexity can go up indefinitely, meaning that some people will always be ahead of AGI and ASI
Bounded Distribution – Humans have a maximum task complexity (Theory of General Relativity)
Outlines 3 Scenarios
1.…
This is machine learning training on landscapes.
StyleGAN3 is a generative adversarial network that is particularly effective at generating photorealistic faces.
A latent space of shapes in generative AI looks like a time machine, because reality is like a time machine.
Millions of years are passing in front of your eyes in a matter of seconds.
Reality is an archetypal film that reproduces a higher dimensional manifold that contains everything at once.…
People entrust their lives to very simple AI mechanisms like traffic lights, I don’t see any reason why public budgets and investments couldn’t be managed by transparent and accountable AI systems.
The real reason for AI regulation is not the “Terminator might come true” argument (only a few ignorant people believe that). The reason is to make it clear to everyone that society needs ‘protectors’ such as corporations and politicians.…
People continue to use LLMs as talking encyclopaedias, completely unaware of the fact that these tools need to be primed with factual data.
…
Terminator not coming yet (sorry AGI maniacs)
New paper by Google provides evidence that transformers (GPT, etc) cannot generalize beyond their training data.
https://arxiv.org/abs/2311.00871
“Our empirical results show transformers demonstrate near-optimal unsupervised model selection capabilities, in their ability to first in-context identify different task families and in-context learn within them when the task families are well-represented in their pretraining data.…
In this insightful talk, Sam Altman emphasises that only a few companies are likely to have the resources to build and maintain Large Language Models (LLMs) such as GPT-3. However, he foresees the emergence of many “layer two” companies with valuations in excess of a billion dollars over the next decade.
“Layer two” is referring to companies built on top of fine-tuned base models that unlock efficiency and progress in domain-specific industries.…
An interesting take on a very interesting topic: can we get creative, valuable ideas from generative AI?.
Some weeks ago I wrote about the core of this problem: finding “what we don´t know that we don´t know“.
But could AI generate the next billion-dollar business idea? Product ideation is the best place to start. If you only have vague initial ideas, generative AI can really help to crystallise them.…
Achieving real value depends on moving from theoretical models to production-level accuracy. This shift requires investment in data tagging and development.
Specialised, fine-tuned models consistently outperform generic counterparts such as ChatGPT, and outperform alternative approaches such as zero-shot learning and prompt-based methods, as shown in studies.…
If you want to get a clearer view of any subject than you have at present, address yourself mentally to the abstract soul of that subject, and ask it to tell you about itself, and you will find that it will do so.
I do not say that it will do this in any miraculous manner, but what you already know of the subject will range itself into a clearer order, and you will see connections that have not previously occurred to you.…
Leyendo cada día sobre inteligencia artificial se diría que los robots vienen a por todos nosotros. Primero por los puestos de trabajo, y luego por las personas.
Los tecno-hippies se han escapado otra vez de Sillicon Valley. Y con sus amigos periodistas le explican a la gente cómo serán sus no-vidas dentro unos pocos años. Siempre muy pocos años, ya que el fin del mundo siempre está muy cerca, pero al mismo tiempo se demora paulatinamente.…
The fact that OpenAI has just released enough updates to destroy a lot of AI startups and plugins should make everyone think.
Does it still make sense to build products on top of the platform? Why should you invest time and money in creating new products that could soon be cannibalised by these people?
Yet it does make sense why? because most users wont get into paid options, not to mention complex configuration menus.…