"É preciso perceber exatamente como a redução da dívida foi feita", diz João Moreira Rato

[MarcoAntonio]

Uma série de artigos sobre IA hoje no Negócios:


IA: O “cavalo selvagem” que veio para ficar

Se a inteligência artificial é uma ferramenta ótima para retirar tarefas repetitivas, dar velocidade aos processos e reduzir as margens de erro, também pode ser um “cavalo selvagem”.

Nuno Boavida: "O risco é termos preparado as pessoas para um mundo que já não existe”

O setor automóvel português tem reforçado a investigação e a aplicação de inteligência artificial sem evidências de redução de emprego. Na banca, que pode entregar “quase toda a gestão de crédito” à inteligência artificial, a transformação foi deliberadamente atrasada por “razões sociais”. Nuno Boavida, que investiga desde 2019 o impacto da inteligência artificial no emprego em Portugal, conta o que tem visto nas empresas. Embora acredite que haverá tempo para requalificar as pessoas, também não afasta o cenário de desemprego em massa ou de desigualdade “insustentável”.

IA: Mármore em piloto pioneiro para extrair vantagem competitiva

Um digitalizador de blocos de mármore que, através de IA, vai prever o interior antes do corte promete abalar a indústria da pedra natural. Projeto descrito como raro no mundo deve partir para a escala industrial em 2026.

Inteligência artificial: os desafios para regular uma tecnologia disruptiva

Adolfo Mesquita Nunes alerta que a regulação para a inteligência artificial não deixa nenhuma empresa de fora, pública e privada. E, apesar das multas pesadas, para o advogado da Pérez-Llorca os maiores riscos são reputacionais.

A singularidade e a exclusão que a IA desencadeia

A evolução permanente e acelerada da IA irá interferir e mudar quase todas as dimensões da vida humana. No entanto, apesar de tomados por esta vertigem, de modo voluntário ou forçado, não devemos esquecer, de modo algum, que, enquanto a todo o segundo cativam para si vastos segmentos de utilizadores, os processos e produtos de IA são agentes de algumas exclusões fundamentais.

[Àlvaro]

A Inteligência artificial é agora o motor da subida, mas a dívida americana um dia irá mostrar que não haverá inteligência que lhe valha. Talvez aguente até às eleições, ou talvez não. Um dia estoira, qual o dia? Isso aí talvez a IA saiba...
Precisamente por causa da dívida é que a conversa de baixar os juros me parece conversa fiada. Baixar a inflação com essa dívida fiscal? Seria absurdo.
É, aliás, esse risco que trouxe novo folego às cripto moedas, que chutou o ouro e a Nvidia (a menos perigosa pelo dinheiro que faz).

[MarcoAntonio]

Startup de Elon Musk capta seis mil milhões em corrida à inteligência artificial

A xAI, startup de Elon Musk, levantou seis mil milhões de dólares para apostar no desenvolvimento de ferramentas de inteligência artificial. Must promete "mais novidades nas próximas semanas".

[MarcoAntonio]

Google scrambles to manually remove weird AI answers in search

/ The company confirmed it is ‘taking swift action’ to remove some of the AI tool’s bizarre responses.

Social media is abuzz with examples of Google’s new AI Overview product saying weird stuff, from telling users to put glue on their pizza to suggesting they eat rocks. The messy rollout means Google is racing to manually disable AI Overviews for specific searches as various memes get posted, which is why users are seeing so many of them disappear shortly after being posted to social networks.

(...)

“A company once known for being at the cutting edge and shipping high-quality stuff is now known for low-quality output that’s getting meme’d,” one AI founder, who wished to remain anonymous, told The Verge.

(...)

Gary Marcus, an AI expert and an emeritus professor of neural science at New York University, told The Verge that a lot of AI companies are “selling dreams” that this tech will go from 80 percent correct to 100 percent. Achieving the initial 80 percent is relatively straightforward since it involves approximating a large amount of human data, Marcus said, but the final 20 percent is extremely challenging. In fact, Marcus thinks that last 20 percent might be the hardest thing of all.

“You actually need to do some reasoning to decide: is this thing plausible? Is this source legitimate? You have to do things like a human fact checker might do, that actually might require artificial general intelligence,” Marcus said. And Marcus and Meta’s AI chief Yann LeCun both agree that the large language models that power current AI systems like Google’s Gemini and OpenAI’s GPT-4 will not be what creates AGI.


(...)

The Verge

[MarcoAntonio]

O autor do FT chama-lhe "halucination" mas os exemplos dados não são de alucinação. Se as respostas são extraídas de fontes que realmente existem (mas cujo intuito era humorístico) então é mais um exemplo de "misalignment".

[MarcoAntonio]

Entretanto, esta cena é deliciosa. E não estou a falar da pizza...

Google promised a better search experience — now it’s telling us to put glue on our pizza

/ Google’s AI Overviews feature is delivering incorrect answers faster than ever.


Imagine this: you’ve carved out an evening to unwind and decide to make a homemade pizza. You assemble your pie, throw it in the oven, and are excited to start eating. But once you get ready to take a bite of your oily creation, you run into a problem — the cheese falls right off. Frustrated, you turn to Google for a solution.

“Add some glue,” Google answers. “Mix about 1/8 cup of Elmer’s glue in with the sauce. Non-toxic glue will work.”

So, yeah, don’t do that. As of writing this, though, that’s what Google’s new AI Overviews feature will tell you to do. The feature, while not triggered for every query, scans the web and drums up an AI-generated response. The answer received for the pizza glue query appears to be based on a comment from a user named “fucksmith” in a more than decade-old Reddit thread, and they’re clearly joking.

(...)

The Verge


O mesmo tema no FT:

Google’s AI search tool tells users to ‘eat rocks’ for your health

‘Hallucinations’ in new feature provoke ridicule and raise questions about betting core business on an experimental product

Google’s new artificial intelligence search tool has advised users that eating rocks can be healthy and to glue cheese to pizza, prompting ridicule and raising questions about its decision to embed an experimental feature into its core product.

“Eating the right rocks can be good for you because they contain minerals that are important for your body’s health,” Google’s AI Overview responded to a query from the Financial Times on Friday, apparently in reference to an April 2021 satirical article from The Onion headlined “Geologists recommend eating at least one small rock per day”.

(...)

FT


Top 10 dos LLM via blind testing. O Gemini 1.5 Pro (Google) colou-se ao ChatGPT 4 Turbo (OpenAI) mas entretantou saiu o 4o.

[MarcoAntonio]

O filme Her, de 2013, parece que foi ontem.

Era Sci-fi. Agora já não!...

Scarlett Johansson, the Hollywood star taking on OpenAI

Going head to head with a powerful company is a role the actor has played before


In the 2013 film Her, a lonely, soon-to-be divorced Theodore Twombly (Joaquin Phoenix) downloads a new AI-powered operating system that its creator, Elements Software, advertises as “an intuitive entity that listens to you, understands you and knows you”.

It is exactly what the despondent Twombly is craving. Settling in front of his home computer, he selects a female voice for his digital assistant and is immediately greeted by a bubbly alto tinged with an irresistible rasp. “Hello, I’m here,” says Samantha, voiced by the actor Scarlett Johansson. He is smitten, and the two soon begin a romantic relationship.

Samantha’s voice appears to have also left a deep impression on Sam Altman, the chief executive of OpenAI, an $86bn artificial intelligence start-up whose ambitions seem to be every bit as lofty as those of the fictional Elements Software.

This week, Johansson claimed that after declining Altman’s offer to provide the voice for his company’s ChatGPT chatbot last September, OpenAI pushed ahead with a new version called Sky that featured a voice so similar to hers, even her closest friends could not tell the difference.

“When I heard the released demo, I was shocked, angered and in disbelief that Mr Altman would pursue a voice that sounded so eerily similar to mine,” Johansson wrote in a statement released on Monday.

After receiving legal notices from Johansson’s team, Altman removed the voice from Sky and apologised to the actor, saying the voice was not intended to resemble hers. The shockwaves are still reverberating in Hollywood and beyond.

(...)

FT


Já agora, convém referir que não é inteiramente claro como é que a OpenAI produziu a voz. A OpenAI alega/alegou que a voz era de uma voice actor anonima, uma resposta um tanto ou quanto opaca tendo em conta o tipo de tecnologia envolvida. O que, por coincidência ou não, tem que ver com o meu post anterior...

[MarcoAntonio]

The Foundation Model Transparency Index

Context. The October 2023 Foundation Model Transparency Index scored 10 major foundation developers like Google and OpenAI for their transparency on 100 transparency indicators. The Index showed developers are generally quite opaque: the average score was a 37 and the top score was a 54 out of 100.

Design. To understand how the landscape has changed, we conduct a follow up 6 months later, scoring developers on the same 100 indicators. In contrast to the October 2023 FMTI, we request that developers submit transparency reports to affirmatively disclose their practices for each of the 100 indicators.

Execution. For the May 2024 Index, 14 developers submitted transparency reports that they have validated and approved for public release. Given their disclosures, we have scored each developer to better understand the new status quo for transparency, changes from October 2023, and areas of sustained and systemic opacity for foundation models.

Evolução de Outubro para Maio:

The Foundation Model Transparency Index after 6 months

[MarcoAntonio]

Elon Musk’s xAI secures new backing from Andreessen, Lightspeed, Sequoia and Tribe

Billionaire’s artificial intelligence start-up closes in on $6bn funding round as it seeks more financial firepower

Elon Musk’s xAI has secured new backing from Silicon Valley venture capital giants Lightspeed Venture Partners, Andreessen Horowitz, Sequoia Capital and Tribe Capital, as the tech billionaire closes in on a new funding round valuing the artificial intelligence start-up at $18bn.

The investors have committed to joining xAI’s latest financing in which Musk is seeking to raise close to $6bn, according to people familiar with the negotiations.

However, one investor involved in the round said the Tesla and X chief remains “a few hundred million dollars” short of that target.

Lightspeed, Tribe and Sequoia declined to comment. Andreessen Horowitz and Musk did not respond to a request for comment. (...)

FT

As voltas que isto dá. Recorde-se que Musk fundou a OpenAI, depois saiu, depois participou numa carta aberta que basicamente pedia uma pausa no desenvolvimento da IA - claro que não teve efeito algum! - e agora anda a tentar voltar a ter uma posição "relevante" na corrida.

[MarcoAntonio]

Meta AI chief says large language models will not reach human intelligence

Yann LeCun argues current AI methods are flawed as he pushes for ‘world modelling’ vision for superintelligence


Meta’s artificial intelligence chief said the large language models that power generative AI products such as ChatGPT would never achieve the ability to reason and plan like humans, as he focused instead on a radical alternative approach to create “superintelligence” in machines.

Yann LeCun, chief AI scientist at the social media giant that owns Facebook and Instagram, said LLMs had “very limited understanding of logic . . . do not understand the physical world, do not have persistent memory, cannot reason in any reasonable definition of the term and cannot plan . . . hierarchically”.

In an interview with the Financial Times, he argued against relying on advancing LLMs in the quest to make human-level intelligence, as these models can only answer prompts accurately if they have been fed the right training data and are, therefore, “intrinsically unsafe”.

Instead, he is working to develop an entirely new generation of AI systems that he hopes will power machines with human-level intelligence, although he said this vision could take 10 years to achieve.

Meta has been pouring billions of dollars into developing its own LLMs as generative AI has exploded, aiming to catch up with rival tech groups, including Microsoft-backed OpenAI and Alphabet’s Google.

LeCun runs a team of about 500 staff at Meta’s Fundamental AI Research (Fair) lab. They are working towards creating AI that can develop common sense and learn how the world works in similar ways to humans, in an approach known as “world modelling”.

(...)

FT

[MarcoAntonio]

Funny...


A 13 de Maio:

ChatGPT will be able to talk to you like Scarlett Johansson in Her (TheVerge)

A 20 de Maio:

OpenAI pulls its Scarlett Johansson-like voice for ChatGPT (The Verge)


Bom, parece que saiu um bocado parecido demais...

Já agora, para quem não estiver a acompanhar a história, Scarlett Johansson entretanto veio revelar que Sam Altman a contactou em Setembro para utilizar a voz dela numa próxima iteração, mas que na altura ela acabou por recusar.

[Bar38]

Arup lost $25mn in Hong Kong deepfake video conference scam

UK-based engineering group identified as target of fraud that used digitally cloned CFO to trick staff

UK engineering group Arup lost HK$200mn ($25mn) after fraudsters used a digitally cloned version of a senior manager to order financial transfers during a video conference, the Financial Times has learned.

Hong Kong police previously revealed what is one of the world’s biggest known deepfake scams, but did not identify the company involved. The FT has confirmed it was Arup, which employs about 18,000 people globally and has annual revenues of more than £2bn.

The case highlights the threat posed by deepfakes — hyper-realistic video, audio or other material generated using artificial intelligence — when used by cyber criminals to target companies or governments.

(...)

FT

Onde é que isto vai para… Até os burlões teem aqui uma porta escancarada (IA) com mil e uma opções, haja imaginação…

[MarcoAntonio]

Arup lost $25mn in Hong Kong deepfake video conference scam

UK-based engineering group identified as target of fraud that used digitally cloned CFO to trick staff

UK engineering group Arup lost HK$200mn ($25mn) after fraudsters used a digitally cloned version of a senior manager to order financial transfers during a video conference, the Financial Times has learned.

Hong Kong police previously revealed what is one of the world’s biggest known deepfake scams, but did not identify the company involved. The FT has confirmed it was Arup, which employs about 18,000 people globally and has annual revenues of more than £2bn.

The case highlights the threat posed by deepfakes — hyper-realistic video, audio or other material generated using artificial intelligence — when used by cyber criminals to target companies or governments.

(...)

FT

[MarcoAntonio]

Investimento em inteligência artificial deve ultrapassar 100 milhões em 2024 em Portugal

Além disso, é também esperado que esta tecnologia registe um crescimento anual de 25%, entre 2023 e 2027, revelou a IDC esta quinta-feira no evento Data & Artificial Intelligence, que se realizou no Centro Cultural de Belém.

[BearManBull]

Google and Harvard unveil most detailed ever map of human brain

The data contained some surprises. For example, rather than forming a single connection, pairs of neurons instead have more than 50. “This is kind of like if two houses on a block had 50 separate phone lines connecting them. What’s going on there? Why are they so strongly connected? We don’t know what the function or significance of this phenomenon is yet, we’re going to have to study it further,” he said.

[MarcoAntonio]

OpenAI releases GPT-4o, a faster model that’s free for all ChatGPT users

/ OpenAI launches a new iteration of its GPT-4 product.

OpenAI is launching GPT-4o, an iteration of the GPT-4 model that powers its hallmark product, ChatGPT. The updated model “is much faster” and improves “capabilities across text, vision, and audio,” OpenAI CTO Mira Murati said in a livestream announcement on Monday. It’ll be free for all users, and paid users will continue to “have up to five times the capacity limits” of free users, Murati added.

In a blog post from the company, OpenAI says GPT-4o’s capabilities “will be rolled out iteratively,” but its text and image capabilities will start to roll out today in ChatGPT.

(...)

New features are coming to ChatGPT’s voice mode as part of the new model. The app will be able to act as a Her-like voice assistant, responding in real time and observing the world around you. The current voice mode is more limited, responding to one prompt at a time and working with only what it can hear.

(...)

The Verge

[BearManBull]

Devia-se proibir tudo que seja derivado de AI.

https://9gag.com/gag/aZZWx99

[MarcoAntonio]

Inteligência Artificial levou Katy Perry ao Met Gala. Como identificar uma "deepfake"?

Especialistas apoiam-se nas falhas da Inteligência Artificial para identificar imagens digitalmente criadas, mas pedem uma maior regulação por parte das plataformas.

[MarcoAntonio]

Sam Altman shoots down reports of search engine launch ahead of Google I/O

/ It’s not GPT-5 or a search engine but a secret third thing.

OpenAI CEO Sam Altman shot down a Reuters report that claims OpenAI is launching a search product next Monday, the day before Google’s flagship conference, Google I/O.

Altman posted on X that while there is an OpenAI announcement set for Monday morning, it’s “not gpt-5, not a search engine,” but whatever it is, he says it “feels like magic.” The only details provided by the official OpenAI post are that the launch will have updates on ChatGPT and its latest model, GPT-4.

The Verge reported on Monday that sources say OpenAI has been aggressively trying to poach Google employees for a team that is working hard to ship the product soon, while Bloomberg and The Information have both reported about a search product in development with no fixed launch date.

The Verge

[BearManBull]

Google biopharma?

AlphaFold is accelerating research in nearly every field of biology.

By solving a decades-old scientific challenge, our AI system is helping to solve crucial problems like treatments for disease or breaking down single-use plastics. One day, it might even help unlock the mysteries of how life itself works.


The AlphaFold solution
It took us four years to solve the protein-folding problem. We began work in 2016, almost immediately after AlphaGo’s victory against Lee Sedol, a top international Go player.

AlphaFold was taught by showing it the sequences and structures of around 100,000 known proteins.

CASP organizes a biennial challenge for research groups to test the accuracy of their protein structure predictions against real experimental data.

Teams are given a selection of amino acid sequences for proteins that have had their exact 3D shape mapped but have not yet been released into the public domain. Teams must submit their best predictions to see how close they are to the subsequently revealed structures.

At CASP13 in 2018, AlphaFold came first. At CASP14 in 2020, we presented AlphaFold 2– which demonstrated a level of accuracy so high that the community considered the protein–folding problem solved.

The AlphaFold 2 methods paper has already received more than 20,000 citations in the scientific literature. This puts it in the top 500 most-cited papers of all time, in any field.

In 2024, together with Isomorphic Labs, we introduced AlphaFold 3, which predicts the structure and interactions of all of life’s molecules.

AlphaFold 3 goes beyond proteins to a broad spectrum of biomolecules including DNA, RNA, and even small molecules, also known as ligands, which encompass many drugs. This leap could unlock more transformative science, from developing biorenewable materials and more resilient crops, to accelerating drug design and genomics research.

Enquanto que uns trabalham por um mundo melhor outros pensam que é com manifestações que se obtêm resultados. Infelizmente os créditos vão para os segundos, socialmente ainda se está na idade da pedra.

[MarcoAntonio]

“A IA tem um risco significativo de aumentar as desigualdades”

O economista norte-americano Joseph Stiglitz acredita que a inteligência artificial (IA) terá impactos importantes no mercado laboral, sobretudo em áreas que até agora estavam a salvo da automação.

[MarcoAntonio]

Buffet alerta para fraudes com IA, cujo poder compara ao de armas nucleares

Berkshire Hathaway está reunida em assembleia geral. Holding vendeu mais 10 milhões de ações da Apple mas Buffet garante que a tomou a decisão por razões fiscais e não devido a qualquer visão sobre a gigante tecnológica.

Warren Buffet afirmou neste sábado que o uso da inteligência artificial (IA) para fraudes pode ser a próxima grande indústria em crescimento e comparou esta tecnologia à das armas nucleares.

(...)

[MarcoAntonio]

(hoje é 1 de Maio, não é 1 de Abril!...)

Turns out the Rabbit R1 was just an Android app all along (The Verge)

Since it launched last week, Rabbit’s R1 AI gadget has inspired a lot of questions, starting with “Why isn’t this just an app?”

Well, friends, that’s because it is just an app.

Over at Android Authority, Mishaal Rahman managed to download Rabbit’s launcher APK on a Google Pixel 6A. With a little tweaking, he was able to run the app as if it were on Rabbit’s own device. Using the volume-up key in place of the R1’s single hardware button, he was able to set up an account and start asking it questions, just as if he was using the $199 R1.

(...)

Tenho as minhas dúvidas. Julgo que o smartphone (ou uma evolução do smartphone como os conhecemos hoje) vão absorver a tecnologia. Não são precisos dois ou três dispositivos e coisas deste género têm falhado sistematicamente em vingar. Mas, aqui vai:

Welcome to the AI gadget era

Humane, Rabbit, Brilliant, Meta, and countless other companies are just about to launch AI-first gadgets. AI hardware may not be as big as the smartphone, but it’s going to be big.

Bem, foi mais rápido do que eu pensava...



Já agora, a desculpa do fabricante é que o R1 "corre na cloud" (onde corre um serviço num servidor que não está implementado em Android mas noutro OS). Mas isto é, para usar uma expressão em inglês, BS. Qualquer smartphone acede à cloud e serviria para fazer a mesma coisa. A questão é que o hardware não é preciso para nada, provavelmente não está a acrescentar nada que não seria implementável num smartphone. Mas eu suspeito que se abandonassem o hardware e o Rabbit R1 se limitasse a uma app no Google Play Store, ninguém estava sequer a falar nele e esse é o "problema". A razão porque o aparelho existe.

[MarcoAntonio]

Dona do ChatGPT assina parceria com Financial Times

O acordo com a publicação económica surge pouco mais de um mês após parcerias semelhantes com o francês Le Monde e com o grupo espanhol Prisa Media, que publica títulos como o El País, Cinco Días ou As.

[MarcoAntonio]

Empresas esperam impactos da IA generativa dentro de três anos

A aposta nesta tecnologia continua a crescer, com a maior parte das empresas a antecipar efeitos “transformativos” até 2027, revela um estudo da Deloitte sobre a IA generativa nas companhias. Em Portugal, cenário é idêntico, afirma Hervé Silva.