i like it, but doesn't fully encapsulate soft power / influential power
@alexandr_wang
-
Evaluating intrinsic talent over external signals matters most
By
–
perhaps the most underrated & important skill: being able to evaluate people on their intrinsics (i.e. their intellect, talent, attitude—their true self) VS their extrinsics (i.e. resumé, brands, external signals) i've met maybe 5 people in my whole life who are great at this
-
SFT, RLHF, and Process Supervision: Fine-tuning Techniques Compared
By
–
fine-tuning in this context:
– SFT/BC (supervised fine tuning & behavioral cloning) as a baseline, but is not powerful enough & has limitations
– RLHF/RLAIF as the big cannons
– Process Supervision, a la the recent @openai @hunterlightman paper as the newest hot technique -
Fine-tuning Essential: Beyond Prompting for LLM Product Control
By
–
thesis I am becoming convicted in: prompting alone—atop even the best LLMs—is NOT enough model control to build a compelling, sticky product as with software, tons of the value comes from fine-grained optimizations, and you can't do that with prompts alone you NEED fine tuning
-
Manager Power Incentives: Team Size and Organizational Misalignment
By
–
funny misalignment: – most employees (esp high performers) want SMALLER teams—fewer people to deal with, more work done – most managers want BIGGER teams—more people = more power, more to “manage” you can tell how strong a manager is by how often they ask for more headcount
-
Corporate Incompetence Masked by Obfuscation in America
By
–
the amount of veiled incompetence in corporate America is staggering in the vast majority of corporate America, there is more effort put towards shrouding incompetence in obfuscation than trying to be competent in the first place
-
Algorithmic Innovation: The Next Frontier Beyond Scaling
By
–
biggest questions in ai: – since the transformer, most AI improvement was via scaling up & great engineering
– almost no algorithmic ideas have driven performance, besides RLHF
– future gains will likely require algorithmic innovation how long will they take? who does them? -
Scale AI and Hugging Face advance LLM evaluation SOTA
By
–
We @scale_AI collaborated with @huggingface to drive forward the SOTA on LLM and AI evaluation. See @natolambert
's excellent Twitter thread to learn more! -
Embrace Repetition to Build Great Things
By
–
most worthwhile pursuits boil down to some deeply repetitive—even rote—task building software -> bugfixing
business/sales -> pitch mtgs
teambuilding -> interviews there is no elusive haven of variety the sooner you embrace the repetition, the sooner you'll build great things