News

Yesterday, the AI startup Anthropic published a paper detailing the successful interpretation of the inner workings of a large language model (LLM). LLMs are notoriously opaque — their size, complexity, and numeric representation of human language have hitherto defied explanation — so it’s impossible to understand why inputs lead to outputs. Anthropic used a technique called dictionary learning, leveraging a sparse encoder to isolate specific concepts within its Claude 3 Sonnet model. The technique allowed them to extract millions of features, including specific entities like the Golden Gate Bridge as well as more abstract ideas such as gender bias. They…

I’m excited to share Generational’s inaugural growth & late-stage company briefing with a deep dive on Scale AI, blending analytical rigor with feature writing. Disclaimer: I have a financial interest in Scale. Don’t take this as investment advice.In this deep dive, you’ll learn insights from conversations with Scale’s customers, ex-employees, and competitors. I could do this thanks to Tegus, which centralizes expert calls into a single platform. Nothing beats primary research when it comes to understanding a company. If you’re curious about Tegus, try them out with this link.Scale AI accelerates the development of AI applications through services and software.…

You may think of the places that remain uncharted by maps as very distant or rural areas. In reality, thriving, bustling communities of millions are not captured in any data source.  In fact, places all around the world have yet to be fully mapped, which is a huge problem. With the increase in climate-related disasters, this lack of data poses serious risks to the most vulnerable populations, especially in the global south. Gaps in data mean gaps in service delivery and gaps in disaster response. Here’s how you can join 600,000 volunteers worldwide and help communities put their citizens, roads,…

If you’ve ran off-the-shelf evals for your tasks, you may have found that most don’t work. They barely correlate with application-specific performance and aren’t discriminative enough to use in production. As a result, we could spend weeks and still not have evals that reliably measure how we’re doing on our tasks. To save us some time, I’m sharing some evals I’ve found useful. The goal is to spend less time figuring out evals so we can spend more time shipping to users. We’ll focus on simple, common tasks like classification/extraction, summarization, and translation. (Although classification evals are basic, having a…

from deephaven import ui, agg, empty_tablefrom deephaven.stream.table_publisher import table_publisherfrom deephaven.stream import blink_to_append_onlyfrom deephaven.plot import express as dxfrom deephaven import updateby as ubyfrom deephaven import dtypes as dhtstocks = dx.data.stocks().reverse()def set_bol_properties(fig): fig.update_layout(showlegend=False) fig.update_traces(fill="tonexty", fillcolor='rgba(255,165,0,0.08)')@ui.componentdef line_plot( filtered_source, exchange, window_size, bol_bands): window_size_key = { "5 seconds": ("priceAvg5s", "priceStd5s"), "30 seconds": ("priceAvg30s", "priceStd30s"), "1 minute": ("priceAvg1m", "priceStd1m"), "5 minutes": ("priceAvg5m", "priceStd5m")} bol_bands_key = {"None": None, "80%": 1.282, "90%": 1.645, "95%": 1.960, "99%": 2.576} base_plot = ui.use_memo(lambda: ( dx.line(filtered_source, x="timestamp", y="price", by="exchange" if exchange == "All" else None, unsafe_update_figure=lambda fig: fig.update_traces(opacity=0.4)) ), [filtered_source, exchange]) window_size_avg_key_col = window_size_key[window_size][0] window_size_std_key_col = window_size_key[window_size][1] avg_plot = ui.use_memo(lambda: dx.line(filtered_source, x="timestamp",…

//php echo do_shortcode('[responsivevoice_button voice="US English Male" buttontext="Listen to Post"]') ?> An expected doubling in power consumption by the world’s data centers during the next few years is expected to strain the capacity of electricity suppliers, according to experts who spoke with EE Times. Those power constraints, without improvements in data center efficiency, will potentially impede the expansion of AI. Electricity demand from data centers, AI and cryptocurrency miners will surge by 2026, the Paris-based International Energy Agency (IEA) said in a January report. After consuming an estimated 460 terawatt-hours (TWh) worldwide in 2022, data centers’ total energy intake could more…

Global R&D hub Imec Flanders today announced at its annual Imec Technology Forum in Antwerp (ITF World 2024) that it has raised a €2.5 billion investment to establish a new R&D pilot line for advanced chip technology and systems-on-chip.  Imec is the world's largest independent research centre in nanoelectronics and digital technology. The NanoIC pilot line is part of the EU Chips Act's vision to accelerate innovation in Europe, stimulate economic growth, and strengthen the European chip industry's ecosystem.  The EU Chips Act focuses on four strategic pilot lines, spread across several European member states to bridge the gap between…