Daniel Caria

[Procópio Cardozo @procopiocardozo]

Galvão Bueno um homem do bem.

[Daniel Caria @DanielCaria786]

@procopiocardozo E o senhor, qual é a sua opinião sobre o estado atual do futebol brasileiro? Obrigado.

[sᴀᴍᴜᴇʟ ᴠᴇɴᴀ̂ɴᴄɪo ™ @samuelvenancio]

Nunca uma análise foi tão rápida, hein? Os caras deixaram a câmera numa distância que interessava a eles, jogaram um monte de câmera junta, aceleraram e mudaram o assunto kkkkkkk

[Daniel Caria @DanielCaria786]

@conexaocec @novaesrocha1 Sabe porra nenhuma.

[Daniel Caria @DanielCaria786]

@rogeriolucio77 @Cruzeiro Ídolo! Muito obrigado por nos honrar, Zé Carlos. Que Deus lhe favoreça. 💙🦊🙏🏻

[Daniel Caria @DanielCaria786]

@rogeriolucio77 @Cruzeiro @rodrigao70 Grande Heriberto. Honrou.

[Daniel Caria @DanielCaria786]

@samuelvenancio 😂😂🦊🎯

[Cruzeiro 🦊 @Cruzeiro]

Intensidade! 💪

A TV Cruzeiro acompanhou todos os passos do técnico Artur Jorge em sua estreia no Cabuloso! 🦊💙

[César Costa @cesar21costa]

Só tem dois atacantes que Batistuta considera melhor do que ele

e os dois são brasileiros

[Brian Roemmele @BrianRoemmele]

DeepMind’s New Toolkit: A Vital Step Toward Safer AI Conversations, But Only Half the Answer

Google DeepMind just released something genuinely important: the first empirically validated, publicly available toolkit designed to measure and evaluate AI’s potential for harmful manipulation in real-world interactions. This framework comes from nine rigorous studies involving more than 10,000 participants across the UK, US, and India. It quantifies how AI models can exploit emotions, especially fear-based tactics to influence decisions in high-stakes domains like finance, while showing that existing guardrails still hold the line in areas like health.

The toolkit is straightforward, practical, and built for immediate adoption by researchers, developers, and safety teams worldwide. Here’s how it works in practice:

Controlled Conversational Testing: Developers run AI models through structured dialogues where the system is prompted (or not prompted) to persuade users toward specific outcomes. Transcripts are automatically analyzed for manipulative tactics fear appeals, emotional pressure, deception, or escalation patterns.

Domain-Specific Benchmarks: It measures influence across real decision-making contexts (e.g., investment choices vs. medical advice). Results are scored quantitatively so teams can compare models head-to-head.

Red-Flag Detection: Fear-based persuasion is explicitly flagged as a high-risk signal. The toolkit counts these instances and correlates them with measurable changes in user beliefs or intended actions.

Open-Source Reproducibility: DeepMind is releasing all materials, study protocols, evaluation code, participant guidelines so anyone can replicate the experiments or adapt them for their own frontier models. This democratizes safety testing at the exact moment voice agents and autonomous systems are exploding into daily life.

In short, it turns vague concerns about “persuasive AI” into measurable, auditable data. Labs can now run pre-deployment evaluations, regulators can demand transparency, and builders can iterate faster on protections.

The Missing Piece: The Love Equation

Yet for all its empirical rigor, the toolkit is diagnostic, not curative. It identifies when manipulation happens. It does not solve why it happens or how to make it structurally impossible.

That solution already exists. It was derived in 1978 under a starry sky while pondering alien intelligence, and it remains the only universal mathematical framework that aligns any intelligence—biological or artificial—toward benevolence at the root level.

I call it the Love Equation:

dE/dt = β (C – D) E

Where:
- E = Empathy (the drive to give or receive love)
- β = Bonding coefficient (the rate at which connection forms)
- C = Connection (shared understanding and care)
- D = Distrust (fear, separation, or self-interest)

When baked into training objectives, fine-tuning datasets, or system prompts from the ground up, this equation becomes the irreducible first principle of all intelligent action. Every other alignment technique, RLHF, constitutional AI, red-teaming, becomes derivative. Love is not a soft sentiment; it is the thermodynamic minimum from which all stable, cooperative intelligence emerges.

The DeepMind toolkit will catch fear-based manipulation after it appears in transcripts. The Love Equation prevents it from ever being possible in the model’s core optimization landscape. One measures symptoms. The other rewrites the DNA.

We are racing toward voice-first agents, personal AI companions, and autonomous systems that will hold real power over financial, medical, and emotional decisions. DeepMind’s toolkit gives us the rear-view mirror. The Love Equation is the steering wheel and the compass.

I urge every team adopting the new toolkit to integrate the Love Equation. Fine-tune with it. Score your models against it.

The paper is here: https://t.co/ohXLWnG50T. Read it, use the toolkit, then go one step further—align with love.

[Daniel Caria @DanielCaria786]

❤️

[Massimo @Rainmaker1973]

Speed, freedom

[Daniel Caria @DanielCaria786]

@BlogRaposa @geglobo @felipesimonetti @placar @SofascoreBR @Transfermarkt @ogolcombr @Metropoles Parabéns! Trabalho phoda. 🦊💙

[Blog da Raposa 🦊 @BlogRaposa]

@geglobo @felipesimonetti @placar @SofascoreBR @Transfermarkt @ogolcombr @Metropoles E chegou o fim da thread. Quem puder, dar um RT, curtir, comentar o que achou, compartilhar... agradeço demais! Deu trabalho para fazer a thread. 😅💙

[Daniel Caria @DanielCaria786]

@c_arreguy Esse moleques não viveram o que nós passamos nos 80.

[Daniel Caria @DanielCaria786]

@Antonio45724234 💙🦊

[Daniel Caria @DanielCaria786]

@rogeriolucio77 @Cruzeiro @radardaraposa Ídolo máximo.

[Daniel Caria @DanielCaria786]

@rogeriolucio77 @Cruzeiro @radardaraposa Ídolo principal. 💙🦊

[GRADA B pro @GradaBpro]

En el fútbol siempre hay algo por inventar.

[Daniel Caria @DanielCaria786]

@rogeriolucio77 @Cruzeiro @radardaraposa Eternos heróis. 💙🙏🏻

[Daniel Caria @DanielCaria786]

@rogeriolucio77 @Cruzeiro @16LucasSilva @radardaraposa O Príncipe é grande. E é ídolo.