The computer science behind translating speech from 100 source languages.

Id say ASI by 2029 or sooner, but we re now in uncharted waters.

Amodei: “I think progress really is as fast as people think it is.”

On a broader level, by pushing AI toward more human-like processing, Titans could mean AI that thinks more deeply than humans — challenging our understanding of human uniqueness and our role in an AI-augmented world.

At the heart of Titans’ design is a concerted effort to more closely emulate the functioning of the human brain. While previous models like Transformers introduced the concept of attention—allowing AI to focus on specific, relevant information—Titans takes this several steps further. The new architecture incorporates analogs to human cognitive processes, including short-term memory, long-term memory, and even the ability to “forget” less relevant information. Perhaps most intriguingly, Titans introduces a concept that’s surprisingly human: the ability to prioritize surprising or unexpected information. This mimics the human tendency to more easily remember events that violate our expectations, a feature that could lead to more nuanced and context-aware AI systems.

The key technical innovation in Titans is the introduction of a neural long-term memory module. This component learns to memorize historical context and works in tandem with the attention mechanisms that have become standard in modern AI models. The result is a system that can effectively utilize both immediate context (akin to short-term memory) and broader historical information (long-term memory) when processing data or generating responses.

This is supposedly super close now. The waters are muddied, we are possibly already well into Agi, and work is underway for ASI.

Super-agents could make AI a true replacement for human workers.

Microsoft research blog.

PromptWizard from Microsoft Research is now open source. It is designed to automate and simplify AI prompt optimization, combining iterative LLM feedback with efficient exploration and refinement techniques to create highly effective prompts in minutes.

Experimental model’s record-breaking performance on science and maths tests wows researchers.

AI models, like ChatGPT-4, can simulate human-like reasoning by solving theory of mind tasks. This breakthrough suggests AI’s potential for advanced social interactions but raises ethical concerns about trust and misuse.

Including the color of vehicles and people’s clothing.

According to Meta, memory layers may be the answer to LLM hallucinations as they don’t require huge compute resources at inference time.

Introduction: Hyperthermia is an established adjunct in multimodal cancer treatments, with mechanisms including cell death, immune modulation, and vascular changes. Traditional hyperthermia applications are resource-intensive and often associated with patient morbidity, limiting their clinical accessibility. Gold nanorods (GNRs) offer a precise, minimally invasive alternative by leveraging near-infrared (NIR) light to deliver targeted hyperthermia therapy (THT). THT induces controlled tumor heating, promoting immunogenic cell death (ICD) and modulating the tumor microenvironment (TME) to enhance immune engagement. This study explores the synergistic potential of GNR-mediated THT with immunotherapies in immunogenically ‘cold’ tumors to achieve durable anti-tumor immunity.

Methods: GNRs from Sona Nanotech Inc. were intratumorally injected and activated using NIR light to induce mild hyperthermia (42–48°C) for 5 minutes. Tumor responses were analyzed for cell death pathways and immune modulation. The immunogenic effects of THT were assessed alone and in combination with intratumoral interleukin-2 (i.t. IL-2) or systemic PD-1 immune checkpoint blockade. Immune cell infiltration, gene expression changes, and tumor growth kinetics were evaluated.

Results: THT reduced tumor burden through cell death mechanisms, including upregulated ICD marked by calreticulin exposure within 48 hours. By 48 hours, CD45+ immune cell levels were increased, including increased levels of immunosuppressive M2 macrophages. While THT led to innate immune cell stimulations highlighted by gene expression upregulation in the STING cGAS pathway and enhanced M1 and dendritic cell levels, tumor regrowth was observed within six days post-treatment. To enhance THT’s immunogenic effects, the therapy was combined with intratumoral interleukin-2 (i.t. IL-2) or systemic PD-1 immune checkpoint blockade. Sequential administration of i.t. IL-2 post-THT induced robust CD8+ T-cell infiltration and led to sustained tumor regression in both treated and distant tumors, accompanied by the emergence of memory T cells. However, IL-2-induced immunosuppressive T-reg populations were also sustained to tumor endpoint suggesting that therapy could be further enhanced.