Orson Terrill

AI Engineer from Applied Statistical Learning and ML focused on dynamical physical systems used in high risk/high value situations.

[Orson Terrill @data_wonder]

@AndrewYNg @Speechlab_ai @AI_Fund 4

[Orson Terrill @data_wonder]

@wandb we do benchmarks specific to our space when we're testing improvements in context learning.

[Orson Terrill @data_wonder]

@wandb I do multi-modality physics informed data science applied to mechanical systems. Could this be an interesting tool for me, or am I in the wrong place?

[Orson Terrill @data_wonder]

@ylecun @tamaybes Many, if not most, casual conversations have dialogue where responses are not fully planned ahead, or at all. The words are made to fit what you've already started to say from a thought that was generated to varying degrees of specificity.

[Orson Terrill @data_wonder]

@stephen_wolfram @ilyasut @OpenAI This sort of thing could end up being one the largest step changes in LLMs external to the development teams producing them. I couldn't fall asleep when I saw this on MLST

[Orson Terrill @data_wonder]

@EzerRatchaga @mayemusk True, and... for all its faults it is efficient use of floor space. Maybe bipeds are the best way for generalized objectives within home spaces and similar vs multiple, or modular, smaller objective specific bots/modules.

[Orson Terrill @data_wonder]

@EzerRatchaga @mayemusk But bipedalism sucks. It's not the best design for anything other than showing a distinctly human mode of transport. The returns to innovation drove bipedal evolution. It's an artifact, like the gills on some people's ears.

[Orson Terrill @data_wonder]

@paulbiggar #TWTR has lost its game if this bonefied zinger goes so quietly into the dark.

[Orson Terrill @data_wonder]

@amirtahmasbi @sh_reya If totally constrained you compromise. The less tailored model has decreased time to deployment. Use understanding of the underlying ML, SME driven FeatEng, and build expressive data for the best use of an automl, or combos of automl. Later use your preferreds. do what it takes.

[Orson Terrill @data_wonder]

@dennisosadebe_ somethings do make sense

[Orson Terrill @data_wonder]

@andrewprock @AndrewHill_BBTV @iam_lpettingill so we're obviously normalizing this.

[Orson Terrill @data_wonder]

@lynnelewis20031 @emilymfg @katiedimartin @lynnelewis20031 Their ego is and always was their real passion, and elitism is their key validating metric.

[Orson Terrill @data_wonder]

@pmddomingos Maybe everyone has heard "shallow guessing".

[Orson Terrill @data_wonder]

@Anni_Maan @svpino The armchair thinking is that ML/DS becomes regularly and ubiquitously utilized, then, code that compiles to machine lang, with numerical computation first design, will be the inevitable outcome. The arc of history may be summarized: Fortran, C, then @JuliaLanguage.

[Orson Terrill @data_wonder]

@JFLA66 @amiannieblack @USATODAY Appendix of the same study.

[Orson Terrill @data_wonder]

@JFLA66 @amiannieblack @USATODAY https://t.co/qZo38dI9rl

[Orson Terrill @data_wonder]

@JFLA66 @amiannieblack @USATODAY https://t.co/b5JArVspsA

[Orson Terrill @data_wonder]

@svpino In school I studied applied stats and math. It definitely helps, but I would never argue someone can't deliver and maintain good ML w/o that background. Know your blind spots and defer/refer accordingly. One will find it impossible without CS skills for the relevant future.

[Orson Terrill @data_wonder]

@svpino Python in Rstudio, in a Jupyter Notebook (via Anaconda), computing on Colab Pro, pushed from Powershell in VScode running on an AWS docker image, executed from a shell inside Visual Studio 6.0 with Windows 98 on a Compaq Computer in my mom's basement.