Photo, posted without a caption
Signed Denis

Channel
@govorit_ai
On this record: Growth · Engagement · Reactions · Posts · Cite this entry
855subscribers
-1 since we began measuring on 7 August 2026
Risers and fallers across the register · movement among entries of Under 1,000.
| Telegram ID | -1001413699800 |
|---|---|
| Type | Channel |
| Username | @govorit_ai |
| Created | 28 December 2018 — measured — cross-checked against a third-party dataset (ext.tg_channel) |
| First recorded | 8 August 2026 |
| Last confirmed live | 28 August 2026 |
| Measurements held | 4 |
| Confirmed unchanged | 1 time, most recently 28 August 2026 |
| On Telegram | t.me/govorit_ai |
| Measured (UTC) | Subscribers | Change |
|---|---|---|
| 28 Aug 2026, 10:13 | 855 | -2 |
| 20 Aug 2026, 22:24 | 857 | +1 |
| 8 Aug 2026, 01:18 | 856 | no change |
| 7 Aug 2026, 15:41 | 856 | first reading |
13 posts held, back to 19 January 2021 — the reader has not yet reached the start of this channel’s public history, so older posts may sit further back, unread. Read across 1 pageof Telegram’s post history, 20 posts per page.
Nothing published in the last 30 days. ERR and ER are rolling 30-day measures, so there is nothing to compute — we hold 13 posts for this entry, the most recent from 29 June 2022. An engagement rate over an empty window would be a number about nothing.
39 reactions across 7 posts, in 4 distinct kinds. The most used accounts for 69.2% of them.
| Reaction | Count | Share | Share, drawn |
|---|---|---|---|
| 👍 | 27 | 69.2% | |
| 🔥 | 8 | 20.5% | |
| 🤔 | 3 | 7.69% | |
| 🤯 | 1 | 2.56% |
No sentiment is inferred, and none should be read in. This table is ordered by count and by nothing else. Emoji do not carry stable meaning across languages or communities — 🙏 is thanks in one channel and mourning in another — so we publish which ones were pressed and how often, and pass no judgement on what an audience meant by them.
Precision. Telegram publishes reaction counts per emoji and short-forms each one — 4.34K, 1.2M — so any single kind at or above 1,000 reaches us at three significant figures, and only counts below 1,000 are exact. The shares above are ratios of those figures and carry the same error. This is also why the total here can differ slightly from a reaction total printed elsewhere on the page: both are sums of the same rounded parts, taken over samples with different edges.
Coverage. Reactions were read on 8 of the 13 sampled posts in this sample. Summed by Telegram’s own count on each post — not by adding up the per-emoji breakdown above — those same posts carry 39reactions in total: the kind of figure the paragraph above means by “a reaction total printed elsewhere on the page”.
Measured over the 13 most recent posts we hold, published 19 January 2021 to 29 June 2022, using the newest reading held for each. Telegram Stars are excluded: they are a payment, not a reaction, and they have their own section.
Photo, posted without a caption
Signed Denis
Typical Decoding for Natural Language Generation https://arxiv.org/abs/2202.00666v2 TL;DR Предложили новый способ сэмплирования из языковых моделей, основанный на концепциях теории информации и предположениях о том, как люди оперируют естественным языком в реальной жизни. Суть подхода Как известно, современные языковые модели способны очень хорошо оценивать вероятности текстов, выдавая низкие значения перплексии дл…
👍15🤔3🔥2🤯1
Signed Denis
Photo, posted without a caption
👍3
Signed Denis
Photo, posted without a caption
🔥2
Signed Denis
Photo, posted without a caption
👍1
Signed Denis
Red Teaming Language Models with Language Models https://arxiv.org/abs/2202.03286 TL;DR Предложили оригинальный способ поиска оскорбительного поведения в языковых моделях: благодаря способностям к zero-shot генерации можно создавать «провокационные» контексты с помощью других языковых моделей. Метод позволяет выявлять bias в модели практически без ручного вмешательства, причём как в режиме «вопрос-ответ», так и при …
🔥4👍2
Signed Denis
Photo, posted without a caption
Signed Renat
Результаты Применяемый подход сравнивали с обучением со “стандартными” параметрами, а также с двумя работами конкурентов - 1) с дискретным 2-ступенчатым увеличением длины последовательности (“2-stage CL“) и 2) плавным увеличением размера батча (“Bsz Warmup 45”). Метрики меряли по перплексии на валидации, перплексии на WikiText и accuracy на LAMBADA. Для анализа “нестабильности” обучения предложили метрику “loss ratio…
Signed Renat
Curriculum Learning: A Regularization Method for Efficient and Stable Billion-Scale GPT Model Pre-Training https://arxiv.org/abs/2108.06084 TL;DR Microsoft показали, что возможно делать pre-training больших (1.5B) GPT-модели с большим размером батча (bs 4К), более стабильно (без скачков лосса) и без потери в качестве - если плавно увеличивать максимальную длину последовательности в ходе обучения. Суть подхода Для …
👍5
Signed Renat
Video, posted without a caption
Signed Artem R
Google презентовал свою новую наработку в Conversational AI под названием LaMDA. Blogpost | Video Судя по посту первого автора GShard сеть представляет из себя огромный MoE transformer на сотни миллиардов (триллионы?) параметров. Тренировали скорее всего на своем внутреннем огромном диалоговом датасете, про который гугл упоминал в Meena - The Meena model has 2.6 billion parameters and is trained on 341 GB of text, f…
👍1
Signed Artem R
Всем привет! На этот раз без пейпера: Думаю, в этом канале сидит много талантливых ML инженеров с интересом к NLP — а мы в нашу замечательную AI команду Replika ищем Senior NLP Research Engineer развивать наш open-domain диалог на миллионах пользователей! Все подробности по ссылке — https://www.notion.so/Senior-NLP-Research-Engineer-Replika-fa43826e6e0f4dc1a13e2b69c3c6f6ef Пишите мне напрямую в телеграм @nikitospr…
Signed Nikita
Showing the 12 most recent of 13 posts we hold for @govorit_ai. View and reaction counts are the latest single reading for each post, not a live figure, and a recent post is still accumulating both. A view count marked ≈ was rounded by Telegram before we ever saw it — t.me prints views in full below 1,000 and to three significant figures above, so ≈1,200,000 means somewhere between 1,150,000 and 1,249,999. Unmarked counts are exact. Text is reproduced from the public post preview and truncated for length.
A live page changes as we take new readings, so a citation should name the measurement it is based on, not just the URL. The line below cites the subscriber count as measured 28 August 2026 — this entry's latest reading, not the date you are reading this.
“Говорит AI” (@govorit_ai), 855 subscribers as measured 28 August 2026. Telegram Register, tgregister.com/channel/govorit_ai.
Full measurement history, CC BY 4.0. Every reading this register holds for this entry, not just the latest one, as a dated, downloadable record: CSV · JSON. Free to use with attribution to tgregister.com. Each file carries its own generation timestamp, which is the figure to cite for exactly when the data was retrieved.