Mukesh Dalal

Working on AI and blockchains/crypto. Previously, Chief AI Officer @ Bose and SBD, Chief Scientist @ BAE, Asst. Prof. @ Columbia, startup founder, etc.

[BrainVTA @Brain_VTA]

Target striatal D1/D2 MSNs without transgenic lines
Enhancer-AAV toolbox: AiE0779m & AiE0780m for D1, AiE0452h for D2, plus interneuron enhancers. Cross-species in mice/rats/macaques.
BrainVTA offers these AAVs for labeling, optogenetics & calcium imaging.
https://t.co/gNrkCFTbFm

[Mukesh Dalal @mukdal]

@mteamisloading I’ll be interested!

[Mukesh Dalal @mukdal]

My #1 reason: the unbounded American dream

[John Arnold @johnarnold]

My top 10 list why the US economy is envy of the world:
- Rule of law
- Large domestic market
- Stable currency
- Most developed financial markets
- Entrepreneurial culture
- Stable political system
- Abundant natural resources
- Research universities
- Immigration
- Federalism

[Andrej Karpathy @karpathy]

# automating software engineering

In my mind, automating software engineering will look similar to automating driving. E.g. in self-driving the progression of increasing autonomy and higher abstraction looks something like:

1. first the human performs all driving actions manually
2. then the AI helps keep the lane
3. then it slows for the car ahead
4. then it also does lane changes and takes forks
5. then it also stops at signs/lights and takes turns
6. eventually you take a feature complete solution and grind on the quality until you achieve full self-driving.

There is a progression of the AI doing more and the human doing less, but still providing oversight. In Software engineering, the progression is shaping up similar:

1. first the human writes the code manually
2. then GitHub Copilot autocompletes a few lines
3. then ChatGPT writes chunks of code
4. then you move to larger and larger code diffs (e.g. Cursor copilot++ style, nice demo here https://t.co/u8ueY0mGxZ)
5....
Devin is an impressive demo of what perhaps follows next: coordinating a number of tools that a developer needs to string together to write code: a Terminal, a Browser, a Code editor, etc., and human oversight that moves to increasingly higher level of abstraction.

There is a lot of work not just on the AI part but also the UI/UX part. How does a human provide oversight? What are they looking at? How do they nudge the AI down a different path? How do they debug what went wrong? It is very likely that we will have to change up the code editor, substantially.

In any case, software engineering is on track to change substantially. And it will look a lot more like supervising the automation, while pitching in high-level commands, ideas or progression strategies, in English.

Good luck to the team!

[Andrej Karpathy @karpathy]

The way to think about asking a factual question to an LLM is that it's a bit like asking a person who read about the topic previously, but they are not allowed to reference any material and have to answer just from memory. LLMs are a lot better at memorizing than humans, but the result is still fundamentally just their best attempt at a lossy recollection. That's the default, unless they have tool use functionality (like Perplexity by default, or Browsing in ChatGPT, or etc.)

(Also my personal use case is not so much articles and "world knowledge", but mostly programming stuff, e.g. docs of linux commands, git, bash, numpy, torch, etc.)

[Russell Kaplan @russelljkaplan]

Predictions for the future of software engineering:

[Mukesh Dalal @mukdal]

@jhagel Love this quote. I’d update it a little bit: The meaning of life is to give life a positive meaning.

[Mukesh Dalal @mukdal]

@VivekGRamaswamy , excellent article. I think your argument naturally extends to the “AI question” with a similar conclusion : an AI should be treated as a human if it exhibits true human-like feelings, pain, love, aspirations, etc.

[Vivek Ramaswamy @VivekGRamaswamy]

A fun & not (entirely) cringeworthy throwback from college! 😬

[Mukesh Dalal @mukdal]

@hmkyale @YaleSEAS @dragomir_radev @Yale @YINSedge @YaleCompsci Very sad and shocking news. Drago was a wonderful person and an excellent researcher. Difficult to believe he is no more. He will be greatly missed.

[Mukesh Dalal @mukdal]

@QuantumPhiloso1 @mhidden @DavidDeutschOxf An agreed truth that contradicts empirical observations or logical/rational reasoning, or motivated by bad morality is dangerous!

[Mukesh Dalal @mukdal]

@GaryMarcus Well, they are brittle in fundamentally different ways. Combining them could lead to a nonbrittle reliable AI.

[Mukesh Dalal @mukdal]

@sama 100% agree @sama. However, this would require AI to become more reliable/trusted, my current passion!

[Mukesh Dalal @mukdal]

Note that ChatGPT is currently a productivity tool for experts, since non-experts may be misled by its hallucinations. Future versions (based on GPT4, etc.) and integrations (Bing, etc.) may remove these limitations.

[Mukesh Dalal @mukdal]

Can ChatGPT + Bing overtake Google Search? Perplexity AI https://t.co/vRz1lBiZ6t

[Mukesh Dalal @mukdal]

@Plinz … that also allows original content.

[Mukesh Dalal @mukdal]

@Plinz Right vision. A key intermediate milestone will be to become a real-time interactive Wikipedia.

[Mukesh Dalal @mukdal]

Uncanny similarity with GPT-3 🙂

[Khai @ThamKhaiMeng]

This is genius. Italian singer Adriano Celentano released a song in the 70s with nonsensical lyrics meant to sound like American English—to prove that Italians would just love any American song. And it was a hit.

[Mukesh Dalal @mukdal]

Interesting images of @BarackObama and @katyperry using stable diffusion with textual inversion concepts: watercolor-portrait and madhubani-art from @huggingface. Incredible progress in Generative AI!