matchagaucho

8 ms·

61.
by matchagaucho 3y ago
YAGNI is a conscious and iterative task prioritization process. From a Pareto perspective, it just means to focus on the 20% of functionality that provides 80% of the value first. That's not to say the other 80% of requests should be i
62.
by matchagaucho 3y ago
These object-oriented "birth projects" were all pretty much deemed failures. NeXTStep by Steve Jobs, C3... Yet all these years later, the principles of OOP, and the companies built on them, outlived the politics that tried to squa
63.
by matchagaucho 3y ago
But AMD was formed to shadow Intel's x86?
64.
by matchagaucho 3y ago
Yes, I noticed the seed/fingerprint caching announcement. Looks promising for all response types.
65.
by matchagaucho 3y ago
Also announced at the OpenAI Developer conference was "JSON mode". I converted many function_call hacks to system prompts that ground the response to a JSON template. Using temperature=0.0 and the keywords "respond using JSON
66.
by matchagaucho 3y ago
Without knowing the target market, I'm thinking the taxonomy of "two-way sync" is understood by a very narrow community. ISVs have been using this approach for years. So maybe that's the audience? If so, then Heroku conn
67.
by matchagaucho 3y ago
Even if this only accomplished one-way directional sync with backup and disaster recovery capabilities, it'd probably find a 10x wider audience.
68.
by matchagaucho 3y ago
If ChatGPT were only 1 LLM, then maybe. But it's a Mixture of Experts (MoE) architecture, which I think makes open source comparisons unfair?
69.
by matchagaucho 3y ago
I like the existing workflows. But I'd really love to GPT-define reverb settings, song forms, channel inputs, ... tedious copy-paste tasks. A local LLM running on x86 or M2 would be great. But API callouts while waiting for the silico
70.
by matchagaucho 3y ago
In a strict "one question / one response" search, raw semantic search results are a great solution. And consumes far fewer tokens. In conversational AI, providing search results appended to a long-memory context produces &quo
71.
by matchagaucho 3y ago
It could also be argued that the Twitter firehose requires substantial RLHF, de-biasing and moderation controls because of its colloquial nature.
72.
by matchagaucho 3y ago
Given that language and text are human inventions to abstractly describe the real-world, it's no wonder the early AI breakthroughs are in text. When a self-driving car gets into an accident, that is a failure of visual/imagery AI,
73.
by matchagaucho 3y ago
Indeed, great article. My first startup had a Telecom-Italia web project (via HP) around 2005. We possibly overlapped. I've since launched 3 long-running apps on Redis. Five-nines and Fast. Much indebted to your work. [̲̅T̲̅][̲̅h̲̅][̲̅
74.
by matchagaucho 3y ago
I'm also seeing a lot of Graphic/3D Artists incorporating AI prompts into their craft. Generative AI is just one-step in the overall process.
75.
by matchagaucho 3y ago
Yes. This is about the ethics of AI training data sources. Peter is OK with AI consuming his tracks. But other Artists... not so much.
76.
by matchagaucho 3y ago
When AI Chess started kicking my @ss, I assumed that human competitive Chess was over. Quite the opposite happened. We appreciate human-vs-human Chess even more because Stockfish, and other AIs, demonstrated there are more challenging lin
77.
by matchagaucho 4y ago
AutoGPT reminds me of the promise of SOA (Service Oriented Architecture) many years ago. The potential is certainly there. But agents need API endpoints and auth credentials to deliver on value.
78.
by matchagaucho 4y ago
Chunk the PDF text and create embeddings. Get cosine similarity between user query and each chunk, and send the top N chunks to OpenAI that fit within token memory.
79.
by matchagaucho 4y ago
he would have easily hit the token limit That was my first question. Do all these tasks fit within a 4K or 8K buffer? Wouldn't be surprised, though, if it works in a 32K GPT4 token limit. Amazing things are possible.
80.
by matchagaucho 4y ago
I wonder how much the scarcity and cost of Nvidia GPUs is driving this message? Nvidia is in a perfect "Arms Dealer" situation right now. Wouldn't be surprised to see the next exponential leap in AI models trained on in-house
81.
by matchagaucho 4y ago
It is marginally better than ChatGPT, though, for writing. I would happily pay a token or credit-based fee to use this while in heavy writing mode. A MRR or ARR license for this? hmmmm... it would need to be multi-modal with dynamic page la
82.
by matchagaucho 4y ago
An LLM running locally providing type-ahead completions seems inevitable.
83.
by matchagaucho 4y ago
In bad times (like covid/recessions), SVB has extended lines of credit to many VC-backed startups. It's been a mutually beneficial relationship for 40 years.
84.
by matchagaucho 4y ago
IIRC there's still a limit of 2,000 shareholders for private companies. So if those employee options transfer to outside buyers, they potentially hit the 2K limit really fast. btw - Nearly this same set of circumstances forced the FB I
85.
by matchagaucho 4y ago
Generating deepfakes of politicians before elections are just another example of the many applications of AI, but not necessarily the fault of the tool. [1] California laws prohibiting deepfakes https://www.dwt.com/insigh
86.
by matchagaucho 4y ago
Most companies will go through 1-2 agile boilerplates before defining, building and owning their own process. If a team moves onto a 3rd boilerplate, they've missed the premise of agile retrospectives.
87.
by matchagaucho 4y ago
Tech Co-ops backed by crypto contracts representing fractional equity ownership are emerging. They are a logical evolution of the Digital Nomad movement from years past (some emerged from Upwork). However, each member is expected to be an i
88.
by matchagaucho 4y ago
as you inevitably have idle workers waiting for new messages. This is no longer the case using the latest in serverless IaaS architecture, where the pool of queue workers dynamically expand to meet demand.
89.
by matchagaucho 4y ago
I find it amusing that some people expect factual responses from ChatGPT. Meanwhile, my script writing friends are having a heyday, and Hollywood is set to green light several new all GPT-enabled scripts. Ad copywriting is forever changed.
90.
by matchagaucho 4y ago
It seems to be the equivalent of right-click "View Source" for webpage HTML/JS source. One view is there's isn't much point in hiding the seed of a dialogue. Another view is if( completion.contains( seedPrompt )

More ›