Semanlink - [2104.10809] Provable Limitations of Acquiring Meaning from Ungrounded Form: What will Future Language Models Understand?

[2104.10809] Provable Limitations of Acquiring Meaning from Ungrounded Form: What will Future Language Models Understand?

Tags:

Au sujet de ce document

sl:arxiv_author :
sl:arxiv_firstAuthor : William Merrill
sl:arxiv_num : 2104.10809
sl:arxiv_published : 2021-04-22T01:00:17Z
sl:arxiv_summary : Language models trained on billions of tokens have recently led to unprecedented results on many NLP tasks. This success raises the question of whether, in principle, a system can ever \"understand\" raw text without access to some form of grounding. We formally investigate the abilities of ungrounded systems to acquire meaning. Our analysis focuses on the role of \"assertions\": contexts within raw text that provide indirect clues about underlying semantics. We study whether assertions enable a system to emulate representations preserving semantic relations like equivalence. We find that assertions enable semantic emulation if all expressions in the language are referentially transparent. However, if the language uses non-transparent patterns like variable binding, we show that emulation can become an uncomputable problem. Finally, we discuss differences between our formal model and natural language, exploring how our results generalize to a modal setting and other semantic relations. Together, our results suggest that assertions in code or language do not provide sufficient signal to fully emulate semantic representations. We formalize ways in which ungrounded language models appear to be fundamentally limited in their ability to \"understand\".@en
sl:arxiv_title : Provable Limitations of Acquiring Meaning from Ungrounded Form: What will Future Language Models Understand?@en
sl:arxiv_updated : 2021-04-22T01:00:17Z
sl:bookmarkOf : https://arxiv.org/abs/2104.10809
sl:creationDate : 2021-05-23
sl:creationTime : 2021-05-23T01:20:07Z

Infos sur le fichier

Bookmark of: https://arxiv.org/abs/2104.10809

Documents with similar tags (experimental)

[2305.12517] Retrieving Texts based on Abstract Descriptions

Tags:

2023-06-15 A propos

Some remarks on Large Language Models

Tags:

> There turned out to be a phase shift somewhere between 60B parameters and 175B parameters, that made language models super impressive.

> **The performance of current days language models are not obtained by language modeling**
>
>    - [Traditional] LMs are not [grounded](tag:grounded_language_learning)
> 
> **3 conceptual steps between GPT-3 and chatGPT: Instructions, code, RLHF.** The last one is, I think, the least interesting despite getting the most attention
>
> Instruction tuning: For example, the human annotators would write something like "please summarize this text", followed by some text they got, followed by a summary they produced of this text. -> Some symbols ("summarize", "translate", "formal") are used in a consistent way together with the concept/task they denote. And they always appear in the beginning of the text. -> the act of producing a summary grounded to the human concept of "summary"
>
> code: programming language code data, and specifically data that contains both natural language instructions or descriptions (in the form of code comments) and the corresponding programming language code. This produced another very direct form of grounding. the human language describes concepts (or intents), which are then realized in the form of the corresponding programs.
>
> "[RL with Human Feedback](tag:reinforcement_learning_from_human_feedback)". This is a fancy way of saying that the model now observes two humans in a conversation, one playing the role of a user, and another playing the role of "the AI", demonstrating how the AI should respond in different situations. This clearly helps the model learn how dialogs work, and how to keep track of information across dialog states (something that is very hard to learn from just "found" data).

2023-01-03 A propos

[2106.10199] BitFit: Simple Parameter-efficient Fine-tuning for Transformer-based Masked Language-models

Tags:

2022-09-01 A propos

[2106.04612] Neural Extractive Search

Tags:

how to extend a
search paradigm we call “**extractive search**” with
neural similarity techniques.

> some information needs require extracting
and aggregating sub-sentence information
(words, phrases, or entities) from multiple documents
(e.g. a list of all the risk factors for a specific
disease and their number of mentions, or a comprehensive
table of startups and CEOs).

> extractive search combines
document selection with information extraction. **The query is extended with capture slots**:
these are **search terms that act as variables, whose
values should be extracted**.
> The user
is then presented with the matched documents, each
annotated with the corresponding captured spans,
as well as aggregate information over the captured
spans

Conclusion :

> We presented a system for neural extractive search.
While we found our system to be useful for scientific
search, it also has clear limitations and areas
for improvement, both in terms of accuracy (only
72.2% of the returned results are relevant, both the
alignment and similarity models generalize well to
some relations but not to others), and in terms of
scale

[Video of demo](https://www.youtube.com/watch?v=TtqWi2GgB5A&t=1832s)

2021-06-23 A propos

[1909.04164] Knowledge Enhanced Contextual Word Representations

Tags:

2020-05-13 A propos

[2004.10151] Experience Grounds Language

Tags:

2020-04-22 A propos

[1608.05426] A Strong Baseline for Learning Cross-Lingual Word Embeddings from Sentence Alignments

Tags:

2018-07-23 A propos